Abstract
Purpose of review
In recent years, legged robots locomotion has been transitioning from mostly flat ground in controlled settings to generic indoor and outdoor environments, approaching now real industrial scenarios. This paper aims at documenting some of the key progress made in legged locomotion control that enabled this transition.
Recent findings
Legged locomotion control makes extensive use of numerical trajectory optimization and its online implementation, model predictive control. A key progress has been how this optimization is handled, with refined models and refined numerical methods. This led the legged locomotion research community to heavily invest in and contribute to the development of new optimization methods and efficient numerical software.
Summary
We present an overview of the typical approach to legged locomotion control, which involves primarily planning a sequence of contacts with the environment and computing a corresponding dynamically feasible trajectory of the Center of Mass of the robot. We then detail recent progress in contact planning and trajectory optimization with either the full Lagrangian dynamics of legged robots or with reduced models.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Introduction
Legs are used by robots to push on their environment with end effectors and provide an active suspension [1], allowing the motion of the main body of the robot to be smoother than the underlying terrain profile. We will colloquially refer to all end effectors used to push on the environment as feet, whatever their form and function. A leg can be temporarily lifted off the ground to make a step, so that discontinuous terrains can be overcome, allowing locomotion in places out of reach otherwise. The number of legs, their kinematic structure and degrees of freedom, and the type of feet (flat, deformable, pointy, wheeled) can vary but the underlying principles are the same. As a matter of fact, the last few years have seen approaches originally developed for biped humanoid robots with flat feet quickly adapted to quadruped robots and all forms of feet.
The control objectives are to find adequate positions for establishing contact between the feet and the environment, and execute a corresponding dynamically feasible motion. The general approach which permeates almost all forms of legged robots locomotion now makes extensive use of numerical trajectory optimization and its online implementation, model predictive control (MPC), in order to handle alternating contacts and the corresponding constraints on contact forces, which are the key difficulty in legged locomotion control.
Aside from adapting this general approach to different numbers of legs and different types of feet, the main progress over the last few years in the field of legged robots locomotion control has been how this trajectory optimization is handled, with refined models and refined numerical methods. This has enabled transitioning from biped walking on mostly flat ground to more general multi-contact locomotion on uneven terrain with various legged robot morphologies.
Typical Approach to Legged Locomotion Control
The Dynamics of Legged Locomotion
The Newton equation of motion of a robot makes it clear that external forces fi generated through contact with the environment are needed to move its center of mass (CoM) c in a direction other than that of gravity g:
where m is the total mass of the robot. The related Euler equation of motion makes it clear that the positions of the contact points si with respect to the CoM c are critical to controlling the angular momentum L of the robot body around its CoM at the same time:
The artificial synergy synthesis approach proposes to focus on this centroidal dynamics Eqs. 1–2 which is directly bound to contact forces, separately from the precise joint posture of the robot which is more directly bound to joint torques [3], and illustrated in Fig. 1. This approach is central today to legged locomotion control, but in situations where balance and posture must be tightly coordinated, the full Lagrangian dynamics of the complete articulated body is preferred as it naturally embeds the centroidal dynamics [4, 5].
One challenge of legged locomotion is that contact forces are usually unilateral: the robot can push on contact surfaces but not pull. Consequently, the forces fi can be oriented only in specific directions, constrained by the limits of friction. Additionally, each contact is binary: either there is contact and a contact force, or there is no contact and no contact force. Impacts can also occur when a leg collides with a surface. Contact switching and impacts are discrete events affecting the continuous time dynamics Eqs. 1–2: the dynamics of legged robots is hybrid in this respect, which constitutes a second challenge.
It is possible in theory to define the set of viable states, from which the robot is able to summon appropriate contact forces to avoid falling. Cyclic motions and equilibrium points are easy to identify as viable, and if the robot is able to reach such a cycle or equilibrium in a few steps from a given state, then it is viable as well [6]. This is the essence of the capturability analysis [7], which is key to many of the existing approaches to legged locomotion control [8].
Control Architecture
A sequence of contact points si is usually planned beforehand in both space and time using random sampling methods, considering the environment of the robot and its goal, accounting for kinematic and static balance constraints [9,10,11]. This is called contact planning. Pre-planned steps can be adjusted later during execution, to adapt to unstable terrain and perturbations [12]. A corresponding CoM motion and angular momentum can be obtained online with a model predictive control scheme [13, 14], accounting for the centroidal dynamics Eqs. 1–2 and making sure that the state of the robot is always kept capturable [8].
When considering locomotion on a flat ground, with co-planar contact points si, the forces fi are typically reduced to their Zero-tilting Moment Point (ZMP), which is bound to the convex hull of contact points [6, 16] (non co-planar contact points can be considered as well with a polyhedral projection of unilateral contact constraints [17]). In this case, efficient linear formulations are possible by pre-defining the vertical motion of the CoM. The most common is the linear inverted pendulum (LIP) model, which assumes a planar CoM motion and zero angular momentum [18]. Accounting for a non-zero angular momentum does not affect the linearity of this model but has rarely been considered [19], probably due to limited benefits when walking on flat ground [20].
The execution of the computed CoM motion, angular momentum and contact positions with the whole body of the robot is usually handled with inverse kinematics [21] or feedback linearization of the whole-body dynamics in the Cartesian space [22], using Quadratic Programs (QPs) to account for instantaneous kinematic, dynamic, force and torque constraints [22,23,24]. Hierarchical QPs are sometimes considered to account for different priorities in the control objectives [25], but singularities can be an issue [26]. Inverse kinematics and feedback linearization can be very sensitive to model inaccuracies, especially on the position, steadiness and stiffness of the ground contact, which are crucial elements for the balance of the robot. Passivity-based approaches are a promising alternative, much less dependent on accurate contact models [27,28,29].
This general approach to legged locomotion control, illustrated in Fig. 2, has been developed originally for biped walking and demonstrated successfully in a wide array of humanoid robots including Kawada’s HRP-2 [13], Honda’s Asimo [30], Aldebaran’s Nao [31] and Boston Dynamics’ Atlas [32••]. More recently, it has been adapted to quadruped robots such as ANYbotics’ ANYmal [33], IIT’s HyQ [34] and MIT’s Cheetah [35], and robots with feet rolling on the ground such as Aldebaran’s Pepper [36] or ETHZ’s wheeled ANYmal [37] and Ascento [38].
Recent Progress in Contact Planning
Contact planning is a particular instance of motion planning, which is known to be PSPACE-hard in general, and therefore NP-hard [39]. In the case of legged locomotion, the discrete nature of contact states is a particular source of combinatorial explosion, adding to the overall non-convexity of the motion planning problem. In order to tackle this unfavorable computational complexity, mixed-integer formulations have been proposed to exploit state-of-the-art off-the-shelf solvers and find globally optimal sequences of contacts in a matter of seconds or even less, accounting for kinematic and even dynamic feasibility constraints on uneven terrain [40,41,42].
An alternative approach builds on the volume reachable by the feet of the robot to provide a necessary and sufficient condition for the existence of a contact sequence. This way, the motion of the robot can be approached without having to specify the exact contact sequence in the first hand, which reduces dramatically the complexity of the motion planning problem, enabling fast online replanning to adapt to changing environments in a fraction of a second. The exact contact sequence is recovered in a second stage using a variety of heuristics such as maximizing robust quasi-static balance, accounting for kinematic and even dynamic feasibility constraints on uneven terrain [15, 43,44,45]. This multistage approach can be very efficient but the reliance on heuristics can be a source of failure.
Recent Progress with Reduced Models
Divergent Component of Motion
One strength of the LIP model is that it is simple enough to lend itself to thorough mathematical analysis. Notably, the divergent part of its dynamics can be represented by a single component, the divergent component of motion (DCM), rather than the full CoM state (position and velocity). Linear feedback control of this DCM emerged as an efficient balance control strategy for biped locomotion over flat terrain [20], simple enough to be demonstrated on a variety of platforms, including position-controlled [46] and torque-controlled robots [47].
The DCM follows a first-order dynamics with respect to contact forces, simpler than the CoM’s second-order dynamics Eq. 1, which helps to simplify other aspects of the control of legged robots. Three recent progresses come to mind. First, the DCM allows removal of the planar CoM constraint to generate 3D walking trajectories with no change in model dynamics [48]. Second, online step-timing adaption can be included in the trajectory optimization while keeping a small convex problem to solve thanks to a suitable change of variable in the analytical DCM trajectory [49]. Third, the DCM can be used as a viability condition in the MPC problem for advanced stability and feasibility proofs [50].
Centroidal Dynamics
In multi-contact scenarios such as locomotion over uneven terrain or in confined spaces, reducing the model of contact forces to their ZMP is limiting. Approaches based on the full centroidal dynamics Eqs. 1–2 have emerged as a solution to compute a feasible CoM trajectory, angular momentum and corresponding contact forces in generic contact configurations. Contact sequences can even be optimized simultaneously [51].
Formulations exploiting the structure of the trajectory optimization problem such as sparsity and convexity have been developed [41, 44, 52]. Additionally, the inertia of quadruped robots is often approximated with that of a single rigid body [34, 35, 51], and locomotion-specific heuristics can be designed to regularize the underlying nonlinear optimization problem so that it can be solved faster around nominal behaviors [53]. With these improvements, approaches based on the full centroidal dynamics can run online with computation times below 10 ms [54].
The centroidal dynamics does not account for kinematic or actuator torque limits. To partly remedy these limitations, extensions of the contact wrench cone to include torque limits have been proposed [55] or learning-based approaches have been introduced to represent more complex whole-body constraints into the centroidal model itself [56].
Mixed Models
An alternative, intermediate solution between reduced and full dynamic models is to combine the centroidal dynamics Eqs. 1–2 with a kinematic model of the complete articulated body as a way to elaborate dynamically feasible whole-body motions while maintaining moderate computational complexity [57, 58]. With this approach, a locally optimal feedback policy can be computed and updated online in less than 70 ms [59]. The centroidal dynamics can also be combined formally with a whole-body dynamic model using an alternating descent method to account for additional aspects of the full dynamics of the robot [60]. An earlier approach combined in a unified, single QP the instantaneous full dynamics of the robot with a LIP model over a receding horizon to capture the long term locomotion dynamics [61].
Recent Progress with Whole-Body Models
Trajectory Optimization
The complete nonlinear dynamics of the robot can be used to obtain efficient coordinations of limb motions, taking into account all kinematic and dynamic feasibility constraints and objectives such as minimal energy consumption through trajectory optimization strategies such as collocation [62, 63] or multiple-shooting [64] (which bring complementary advantages in terms of speed, accuracy, stability and precision [65]). Contact sequences can even be optimized simultaneously, using relaxations of the contact model [66, 67]. Thanks to efficient open-source numerical solvers [68,69,70] exploiting the sparsity of the trajectory optimization problem and exact derivatives of the dynamic models [71,72,73], whole-body motions can be computed online in less than 7 ms [70, 74, 75].
Reinforcement Learning
The full dynamic model of the robot can also be used offline to compute directly an optimal feedback control policy operating from raw sensor measurements and an estimate of the robot attitude, outperforming more traditional approaches in challenging and difficult to model indoor and outdoor environments thanks to mature reinforcement learning techniques [76, 77••]. Key ingredients for this approach to be effective are as follows: (i) an efficient simulator to generate a large amount of sample behaviors [78], (ii) a curriculum strategy to guide the optimization of the feedback control policy from simple scenarios to more complex ones and (iii) accounting for difficult to model uncertainties in the actuator models using a neural network trained on real hardware measurements.
Conclusions
Legged locomotion means establishing a sequence of contacts between the feet and the environment of the robot, and executing a corresponding kinematically and dynamically feasible motion. Numerical trajectory optimization plays a central role, to handle alternating contacts and the corresponding constraints on contact forces. Models of varying complexity, from the simplest LIP model, to full centroidal dynamics Eq. 1–Eq. 2, to models of the complete articulated body can be used either alone or in combination to handle different aspects of locomotion. Refinements in these models and advanced numerical methods, either off-the-shelf or specifically tailored to legged locomotion, enabled a transition in recent years from flat terrain to generic indoor and outdoor environments. The first successful demonstration of reinforcement learning for quadruped locomotion in real, complex situations was also achieved in the same way.
The central role of numerical optimization, both online and offline, led the legged locomotion research community to heavily invest in and contribute to the development of new optimization methods and efficient numerical software. Notable contributions were made to Lexicographic Programming to safely handle multiple objectives with varying priorities, differential dynamic programming (DDP) to leverage the sparsity of nonlinear trajectory optimization problems, and even combinations of these two approaches [79]. This involved the development of efficient and versatile open-source software for dynamic models [80,81,82], simulation [83], state estimation [84] and control design [85].
Open-source software, and open-source hardware with printable parts, simplified electronics and actuation [86, 87•] could facilitate future reproducibility, dissemination and democratization of well-established, robust and extensively tested solutions, fostering faster iterations and innovations with the possibility to easily share, extend and improve existing implementations (a list is provided in Table 1).
Interestingly, the existing literature on legged locomotion control relies almost exclusively on state feedback, but the position and velocity of the CoM cannot be measured directly as it is a virtual point and contacts with the environment can be difficult to detect and measure. State estimation is therefore a crucial component, which has been surprisingly little explored in this respect [2, 84, 89,90,91,92,93]. The reliance on state estimation and feedback makes existing legged robots extremely dependent on accurate hardware, expensive and brittle. Sensor-based control could be an alternative, largely unexplored so far [94]. Robust control could be another alternative, not much explored either [95,96,97,98]. A recent conclusion from robust control analysis of legged balance is that perfectly stable feedback can be obtained at surprisingly low frequency, as demonstrated with both humanoid [97] and quadruped robots [59].
Building on the open-source software and hardware mentioned earlier, such control approaches could contribute to the development of cheaper, more robust and versatile robots, which looks like the next logical step now that legged locomotion is being demonstrated robustly in generic indoor and outdoor environments and approaching real industrial scenarios [99, 100]. Deformable elements could also contribute to this objective [101].
References
Papers of particular interest, published recently, have been highlighted as: • Of importance •• Of major importance
Waldron KJ, McGhee RB. The adaptive suspension vehicle. IEEE Control Syst Mag 1986;6: 7–12.
Carpentier J, Benallegue M, Mansard N, Laumond J-P. Center of mass estimation for polyarticulated system in contact — a spectral approach. IEEE Trans Robot 2016;32(4):810–22.
Vukobratović MK. Contribution to the study of anthropomorphic systems. Kybernetika 1972; 8(5):404–18.
Wieber P-B. Holonomy and nonholonomy in the dynamics of articulated motion. Proceedings of the Ruperto Carola symposium on fast motion in biomechanics and robotics; 2005.
Orin DE, Goswami A, Lee S-H. Centroidal dynamics of a humanoid robot. Auton Robot 2013;35(2-3):161–76.
Wieber P-B. On the stability of walking systems. Proceedings of the international workshop on humanoids and human friendly robots; 2002.
Pratt J, Tedrake R. Velocity based stability margins for fast bipedal walking. Proceedings of the Ruperto Carola symposium on fast motion in biomechanics and robotics; 2005.
Wieber P-B, Tedrake R, Kuindersma S. Modeling and control of legged robots. Handbook of robotics, chap 48. In: Siciliano B and Khatib O, editors. 2nd edn. Springer; 2016. p. 1203–34.
Bretl T. Motion planning of multi-limbed robots subject to equilibrium constraints: the free-climbing robot problem. Int J Robot Res 2006;25(4):317–42.
Hauser K, Bretl T, Latombe J-C, Harada K, Wilcox B. 2008. Motion planning for legged robots on varied terrain. Int J Robot Res.
Escande A, Kheddar A, Miossec S. Planning contact points for humanoid robots. Robot Auton Syst 2013;61(5):428–42.
Feng S, Xinjilefu X, Atkeson CG, Kim J. Robust dynamic walking using online foot step optimization. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2016.
Kajita S, Kanehiro F, Kaneko K, Fujiwara K, Harada K, Yokoi K, Hirukawa H. Biped walking pattern generation by using preview control of zero moment point. Proceedings of the IEEE international conference on robotics & automation; 2003. p. 1620–26.
Wieber P-B. Trajectory free linear model predictive control for stable walking in the presence of strong perturbations. Proceedings of the IEEE-RAS international conference on humanoid robots; 2006.
Fernbach P, Tonneau S, Stasse O, Carpentier J, Taïx M. c-CROC: Continuous and convex resolution of centroidal dynamic trajectories for legged robots in multicontact scenarios. IEEE Trans Robot 2020;36(3):676–91.
Sardain P, Bessonnet G. Forces acting on a biped robot. Center of pressure—zero moment point. IEEE Trans Sys Man Cybern – Part A 2004;34(5):630–37.
Caron S, Pham Q-C, Nakamura Y. ZMP support areas for multi-contact mobility under frictional constraints. IEEE Trans Robot 2017;33(1):67–80.
Kajita S, Tani K. Study of dynamic biped locomotion on rugged terrain – derivation and application of the linear inverted pendulum mode. Proceedings of the IEEE international conference on robotics & automation; 1991. p. 1405–11.
Seyde T, Shrivastava A, Englsberger J, Bertrand S, Pratt J, Griffin R. Inclusion of angular momentum during planning for capture point based walking. Proceedings of the IEEE international conference on robotics & automation; 2018. p. 1791–98.
Koolen T, de Boer T, Rebula J, Goswami A, Pratt J. Capturability-based analysis and control of legged locomotion, part 1: Theory and application to three simple gait models. Int J Robot Res 2012; 31(9):1094–1113.
Kajita S, Kanehiro F, Kaneko K, Fujiwara K, Harada K, Yokoi K, Hirukawa H. Resolved momentum control: Humanoid motion planning based on the linear and angular momentum. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2003. p. 1644–50.
Wieber P-B. Constrained dynamics and parametrized control in biped walking. Proceedings of the international symposium on mathematical theory of networks and systems; 2000.
Fujimoto Y, Kawamura A. Simulation of an autonomous biped walking robot including environmental force interaction. IEEE Robot Autom Mag 1998;5(2):33–41.
Kuindersma S, Permenter F, Tedrake R. An efficiently solvable quadratic program for stabilizing dynamic locomotion. Proceedings of the IEEE international conference on robotics & automation; 2014. p. 2589–94.
Escande A, Mansard N, Wieber P-B. Hierarchical quadratic programming: Fast online humanoid-robot motion generation. Int J Robot Res 2014;33(7):1006–28.
Wieber P-B, Escande A, Dimitrov D, Sherikov A. Geometric and numerical aspects of redundancy. Geometric and numerical foundations of movements. In: Laumond J-P, Mansard N, and Lasserre J-B, editors. Springer International Publishing; 2017. p. 67–85.
Henze B, Roa MA, Ott C. 2016. Passivity-based whole-body balancing for torque-controlled humanoid robots in multi-contact scenarios. Int J Robot Res.
Mesesan G, Englsberger J, Garofalo G, Ott C, Albu-Schäffer A. Dynamic walking on compliant and uneven terrain using DCM and passivity-based whole-body control. Proceedings of the IEEE-RAS international conference on humanoid robots; 2019.
Kurtz V, Wensing PM, Lin H. 2020. Approximate simulation for template-based whole-body control. IEEE Robot Autom Lett.
Takenaka T, Matsumoto T, Yoshiike T. Real time motion generation and control for biped robot -1st report: Walking gait pattern generation. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2009.
Gouaillier D, Collette C, Kilner C. Omni-directional closed-loop walk for NAO. Proceedings of the IEEE-RAS international conference on humanoid robots; 2010.
•• Kuindersma S. Recent progress on Atlas, the world’s most dynamic humanoid robot. This recorded lecture provides precise details about the complete control design behind Boston Dynamics’ world-leading robot videos. [Online]. Available: https://www.youtube.com/watch?v=EGABAx52GKI.
Bellicoso CD, Jenelten F, Gehring C, Hutter M. Dynamic locomotion through online nonlinear motion optimization for quadrupedal robots. IEEE Robot Autom Lett 2018;3(3):2261–68.
Mastalli C, Focchi M, Havoutis I, Radulescu A, Calinon S, Buchli J, Caldwell DG, Semini C. Trajectory and foothold optimization using low-dimensional models for rough terrain locomotion. Proceedings of the IEEE international conference on robotics & automation; 2017.
Di Carlo J, Wensing PM, Katz B, Bledt G, Kim S. Dynamic locomotion in the MIT Cheetah 3 through convex model predictive control. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2018.
Lafaye J, Gouaillier D, Wieber P-B. Linear model predictive control of the locomotion of Pepper, a humanoid robot with omnidirectional wheels. Proceedings of the IEEE-RAS international conference on humanoid robots; 2014.
Bjelonic M, Sankar PK, Bellicoso CD, Vallery H, Hutter M. Rolling in the deep – hybrid locomotion for wheeled-legged robots using online trajectory optimization. IEEE Robot Autom Lett 2020;5(2): 3626–33.
Klemm V, Morra A, Gulich L, Mannhart D, Rohr D, Kamel M, de Viragh Y, Siegwart R. LQR-Assisted whole-body control of a wheeled bipedal robot with kinematic loops. IEEE Robot Autom Lett 2020;5(2):3745–52.
Lavalle SM. Planning algorithms. Cambridge: Cambridge University Press; 2006.
Deits R, Tedrake R. Footstep planning on uneven terrain with mixed-integer convex optimization. Proceedings of the IEEE-RAS international conference on humanoid robots; 2014.
Ponton B, Herzog A, Schaal S, Righetti L. A convex model of humanoid momentum dynamics for multi-contact motion generation. Proceedings of the IEEE-RAS international conference on humanoid robots; 2016.
Aceituno-Cabezas B, Mastalli C, Dai H, Focchi M, Radulescu A, Caldwell DG, Cappelletto J, Grieco JC, Fernández-López G, Semini C. Simultaneous contact, gait, and motion planning for robust multilegged locomotion via mixed-integer convex optimization. IEEE Robot Autom Lett 2017;3 (3):2531–38.
Perrin N, Ott C, Englsberger J, Stasse O, Lamiraux F, Caldwell DG. Continuous legged locomotion planning. IEEE Trans Robot 2016;33(1):234–39.
Carpentier J, Tonneau S, Naveau M, Stasse O, Mansard N. A versatile and efficient pattern generator for generalized legged locomotion. Proceedings of the IEEE international conference on robotics & automation. IEEE; 2016. p. 3555–61.
Tonneau S, Del Prete A, Pettré J, Park C, Manocha D, Mansard N. An efficient acyclic contact planner for multiped robots. IEEE Trans Robot 2018;34(3):586–601.
Kajita S, Morisawa M, Miura K, Nakaoka S, Harada K, Kaneko K, Kanehiro F, Yokoi K. Biped walking stabilization based on linear inverted pendulum tracking. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2010. p. 4489–96.
Englsberger J, Ott C, Roa MA, Albu-Schäffer A, Hirzinger G. Bipedal walking control based on capture point dynamics. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2011.
Englsberger J, Ott C, Albu-Schaffer A. Three-dimensional bipedal walking control based on divergent component of motion. IEEE Trans Robot 2015;31(2):355–68. [Online]. Available: http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7063218.
Khadiv M, Herzog A, Moosavian SAA, Righetti L. Walking control based on step timing adaptation. IEEE Trans Robot 2020;36(3):629–43. [Online]. Available: https://ieeexplore.ieee.org/document/9082021/.
Scianca N, De Simone D, Lanari L, Oriolo G. 2020. MPC for humanoid gait generation: Stability and feasibility. IEEE Trans Robot 36.
Winkler AW, Bellicoso CD, Hutter M, Buchli J. Gait and trajectory optimization for legged systems through phase-based end-effector parameterization. IEEE Robot Autom Lett 2018;3(3):1560–67.
Herzog A, Schaal S, Righetti L. Structured contact force optimization for kino-dynamic motion generation. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2016. p. 2703–10.
Bledt G, Kim S. Extracting legged locomotion heuristics with regularized predictive control. Proceedings of the IEEE international conference on robotics & automation; 2020. p. 406–12.
Bledt G, Kim S. Implementing regularized predictive control for simultaneous real-time footstep and ground reaction force optimization. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2019. p. 6316–23.
Orsolino R, Focchi M, Caron S, Raiola G, Barasuol V, Caldwell DG, Semini C. Feasible region: an actuation-aware extension of the support region. IEEE Trans Robot 2020;36(4):1239–55.
Carpentier J, Mansard N. Multicontact locomotion of legged robots. IEEE Trans Robot 2018; 34(6):1441–60.
Dai H, Valenzuela A, Tedrake R. Whole-body motion planning with centroidal dynamics and full kinematics. Proceedings of the IEEE-RAS international conference on humanoid robots; 2014. p. 295–302.
Herzog A, Rotella N, Schaal S, Righetti L. Trajectory generation for multi-contact momentum control. Proceedings of the IEEE-RAS international conference on humanoid robots; 2015. p. 874–80.
Grandia R, Farshidian F, Ranftl R, Hutter M. Feedback MPC for torque-controlled legged robots. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2019.
Budhiraja R, Carpentier J, Mansard N. Dynamics consensus between centroidal and whole-body models for locomotion of legged robots. Proceedings of the IEEE international conference on robotics & automation. IEEE; 2019. p. 6727–33.
Sherikov A, Dimitrov D, Wieber P-B. Whole body motion controller with long-term balance constraints. Proceedings of the IEEE-RAS international conference on humanoid robots; 2014. p. 444–50.
Posa M, Kuindersma S, Tedrake R. Optimization and stabilization of trajectories for constrained dynamical systems. Proceedings of the IEEE international conference on robotics & automation; 2016.
Hereid A, Hubicki CM, Cousineau EA, Ames AD. Dynamic humanoid locomotion: a scalable formulation for HZD gait optimization. IEEE Trans Robot 2018;34(2):370–87.
Koch KH, Mombaur K, Stasse O, Souè res P. Optimization based exploitation of the ankle elasticity of HRP-2 for overstepping large obstacles. 2014 IEEE-RAS international conference on humanoid robots. IEEE; 2014. p. 733–40.
Diehl M, Bock HG, Diedam H, Wieber P-B. Fast direct multiple shooting algorithms for optimal robot control. Fast motions in biomechanics and robotics. Springer; 2006. p. 65–93.
Mordatch I, Todorov E, Popović Z. Discovery of complex behaviors through contact-invariant optimization. ACM Trans Graph (TOG) 2012;31(4):1–8.
Posa M, Cantu C, Tedrake R. A direct method for trajectory optimization of rigid bodies through contact. Int J Robot Res 2014;33(1):69–81.
Giftthaler M, Neunert M, Stäuble M, Buchli J. The control toolbox — an open-source C++ library for robotics, optimal and model predictive control. 2018 IEEE international conference on simulation, modeling, and programming for autonomous robots (SIMPAR); 2018. p. 123–29.
Howell TA, Jackson BE, Manchester Z. ALTRO: A fast solver for constrained trajectory optimization. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2019. p. 7674–79.
Mastalli C, Budhiraja R, Merkt W, Saurel G, Hammoud B, Naveau M, Carpentier J, Righetti L, Vijayakumar S, Mansard N. Crocoddyl: an efficient and versatile framework for multi-contact optimal control. Proceedings of the IEEE international conference on robotics & automation; 2020. p. 2536–42.
Giftthaler M, Neunert M, Stäuble M, Frigerio M, Semini C, Buchli J. Automatic differentiation of rigid body dynamics for optimal control and estimation. Adv Robot 2017;31(22):1225–37.
Ayusawa K, Yoshida E. Comprehensive theory of differential kinematics and dynamics towards extensive motion optimization framework. Int J Robot Res 2018;37(13-14):1554–72.
Carpentier J, Mansard N. Analytical derivatives of rigid body dynamics algorithms. Robotics: Science and systems (RSS 2018); 2018.
Koenemann J, Prete AD, Tassa Y, Todorov E, Stasse O, Bennewitz M, Mansard N. Whole-body model-predictive control applied to the HRP-2 humanoid. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2015.
Neunert M, Stäuble M, Giftthaler M, Bellicoso CD, Carius J, Gehring C, Hutter M, Buchli J. Whole-body nonlinear model predictive control through contacts for quadrupeds. IEEE Robot Autom Lett 2018;3(3):1458–65.
Hwangbo J, Lee J, Dosovitskiy A, Bellicoso D, Tsounis V, Koltun V, Hutter M. 2019. Learning agile and dynamic motor skills for legged robots. Sci Robot 4(26).
•• Lee J, Hwangbo J, Wellhausen L, Koltun V, Hutter M. 2020. Learning quadrupedal locomotion over challenging terrain. Sci Robot 5(47).
Hwangbo J, Lee J, Hutter M. Per-contact iteration method for solving contact dynamics. IEEE Robot Autom Lett 2018;3(2):895–902.
Geisert M, Del Prete A, Mansard N, Romano F, Nori F. Regularized hierarchical differential dynamic programming. IEEE Trans Robot 2017;33(4):819–33.
Felis ML. RBDL: An efficient rigid-body dynamics library using recursive algorithms. Auton Robot 2017;41(2):495–511.
Carpentier J, Saurel G, Buondonno G, Mirabel J, Lamiraux F, Stasse O, Mansard N. The Pinocchio C++ library – a fast and flexible implementation of rigid body dynamics algorithms and their analytical derivatives. IEEE international symposium on system integrations (SII); 2019.
Koolen T, Deits R. Julia for robotics: simulation and real-time control in a high-level programming language. Proceedings of the IEEE international conference on robotics & automation; 2019. p. 604–11.
Lee J, Grey MX, Ha S, Kunz T, Jain S, Ye Y, Srinivasa SS, Stilman M, Liu CK. DART: Dynamic animation and robotics toolkit. J Open Sour Softw 2018;3(22):500.
Camurri M, Ramezani M, Nobili S, Fallon M. Pronto: A multi-sensor state estimator for legged robots in real-world scenarios. Front Robot AI 2020;7(68):1–18. [Online]. Available: https://www.frontiersin.org/article/10.3389/frobt.2020.00068.
Tedrake R, the Drake Development Team. 2019. Drake: Model-based design and verification for robotics. [Online]. Available: https://drake.mit.edu.
Lapeyre M, Rouanet P, Grizou J, N’Guyen S, Falher AL, Depraetre F, Oudeyer P. Poppy: Open source 3D printed robot for experiments in developmental robotics. Proceedings of the international conference on development and learning and on epigenetic robotics; 2014.
• Grimminger F, Meduri A, Khadiv M, Viereck J, Wüthrich M, Naveau M, Berenz V, Heim S, Widmaier F, Flayols T, Fiene J, Badri-Spröwitz A, Righetti L. An open torque-controlled modular robot architecture for legged locomotion research. IEEE Robot Autom Lett 2020;5(2):3650–57.
Mirabel J, Tonneau S, Fernbach P, Seppälä A-K, Campana M, Mansard N, Lamiraux F. HPP: A new software for constrained motion planning. 2016 IEEE/RSJ international conference on intelligent robots and systems (IROS). IEEE; 2016. p. 383–89.
Manuelli L, Tedrake R. Localizing external contact using proprioceptive sensors: the contact particle filter. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2016. p. 5062–69.
Blösch M. 2017. State estimation for legged robots - kinematics, inertial sensing and computer vision. Eidgenössische Technische Hochschule zürich PhD thesis.
Flayols T, Del Prete A, Wensing P, Mifsud A, Benallegue M, Stasse O. Experimental evaluation of simple estimators for humanoid robots. Proceedings of the IEEE-RAS international conference on humanoid robots. IEEE; 2017. p. 889–95.
Rotella N. 2017. Estimation-based control for humanoid robots. University of Southern California PhD thesis.
Fourmy M, Flayols T, Mansard N, Solà J. Contact forces pre-integration for the whole body estimation of legged robots. Proceedings of the IEEE international conference on robotics & automation; 2021.
Samson C, Espiau B. Application of the task-function approach to sensor-based control of robot manipulators. IFAC Proc Vol 1990;123(8, Part 5):269–74.
Del Prete A, Mansard N. Robustness to joint-torque-tracking errors in task-space inverse dynamics. IEEE Trans Robot 2016;32(5):1091–105.
Farshidian F, Jelavić E, Winkler AW, Buchli J. Robust whole-body motion control of legged robots. Proceedings of the IEEE/RSJ international conference on intelligent robots & systems; 2017. p. 4589–96.
Villa NA, Englsberger J, Wieber P-B. Sensitivity of legged balance control to uncertainties and sampling period. IEEE Robot Autom Lett 2019;4(4):6.
Hammoud B, Khadiv M, Righetti L. 2021. Impedance optimization for uncertain contact interactions through risk sensitive optimal control. IEEE Robot Autom Lett.
Kheddar A, Caron S, Gergondet P, Comport A, Tanguy A, Ott C, Henze B, Mesesan G, Englsberger J, Roa MA, Wieber P-B, Chaumette F, Spindler F, Oriolo G, Lanari L, Escande A, Chappellet K, Kanehiro F, Rabaté P. Humanoid robots in aircraft manufacturing: the Airbus use cases. IEEE Robot Autom Mag 2019;26(4):30–45.
Gehring C, Fankhauser P, Isler L, Diethelm R, Bachmann S, Potz M, Gerstenberg L, Hutter M. 2021. ANYMal in the field: Solving industrial inspection of an offshore HVDC platform with a quadrupedal robot. Field Serv Robot 247–60.
Werner A, Henze B, Loeffl F, Leyendecker S, Ott C. Optimal and robust walking using intrinsic properties of a series-elastic robot. Proceedings of the IEEE-RAS international conference on humanoid robots; 2017.
Acknowledgements
The authors would like to thank Stéphane Caron for discussions that helped shape the present document.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of Interest
The authors declare no competing interests.
Additional information
Human and Animal Rights and Informed Consent
This article does not contain any studies with human or animal subjects performed by any of the authors.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This article belongs to the Topical Collection on Humanoid and Bipedal Robotics
Rights and permissions
About this article
Cite this article
Carpentier, J., Wieber, PB. Recent Progress in Legged Robots Locomotion Control. Curr Robot Rep 2, 231–238 (2021). https://doi.org/10.1007/s43154-021-00059-0
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s43154-021-00059-0