Direction-changing fall control of humanoid robots: theory and experiments

Goswami, Ambarish; Yun, Seung-kook; Nagarajan, Umashankar; Lee, Sung‐Hee; Yin, KangKang; Kalyanakrishnan, Shivaram

doi:10.1007/s10514-013-9343-2

Cited by 36 publications

(17 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To accelerate the computation, various simplified models have been proposed to approximate falling motions, such as an inverted pendulum [7], [8], a planar robot in the sagittal plane [9], a tripod [10], and a sequence of inverted pendulums [1]. In spite of the effort to reduce the computation, most of the optimization-based techniques are still too slow for real-time applications, with the exception of the work done by Goswami et al [11], who proposed to compute the optimal stepping location to change the falling direction. In contrast, our work takes the approach of policy learning using deep reinforcement learning techniques.…”

Section: Related Workmentioning

confidence: 99%

Learning a unified control policy for safe falling

Visak

Liu

2017

2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

Being able to fall safely is a necessary motor skill for humanoids performing highly dynamic tasks, such as running and jumping. We propose a new method to learn a policy that minimizes the maximal impulse during the fall. The optimization solves for both a discrete contact planning problem and a continuous optimal control problem. Once trained, the policy can compute the optimal next contacting body part (e.g. left foot, right foot, or hands), contact location and timing, and the required joint actuation. We represent the policy as a mixture of actor-critic neural network, which consists of n control policies and the corresponding value functions. Each pair of actor-critic is associated with one of the n possible contacting body parts. During execution, the policy corresponding to the highest value function will be executed while the associated body part will be the next contact with the ground. With this mixture of actor-critic architecture, the discrete contact sequence planning is solved through the selection of the best critics while the continuous control problem is solved by the optimization of actors. We show that our policy can achieve comparable, sometimes even higher, rewards than a recursive search of the action space using dynamic programming, while enjoying 50 to 400 times of speed gain during online execution.

show abstract

Section: Related Workmentioning

confidence: 99%

Learning a unified control policy for safe falling

Visak

Liu

2017

2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

show abstract

“…Only three studies have been reported for biped robots [4], [17], [28]. The focus for these studies were prediction of a fall before it occurs and taking corrective measures to prevent the fall.…”

Section: Related Workmentioning

confidence: 99%

Semi-Automatic Extraction of Training Examples From Sensor Readings for Fall Detection and Posture Monitoring

et al. 2016

View full text Add to dashboard Cite

While inexpensive wearable motion-sensing devices have shown great promise for fall detection and posture monitoring, two major problems still exist and have to be solved: 1) a framework for the development of firmware and 2) software to make intelligent decisions. We address both the problems. We propose a generic framework for developing the firmware. We also demonstrate that the k-means clustering algorithm can semi-automatically extract training examples from motion data. Moreover, we trained and evaluated several one-and two-level classification networks to monitor non-fall activities and to detect fall events. The proposed classification networks are the combinations of neural networks and softmax regression. These networks are trained offline with examples extracted by our proposed method. The cross-validation of trained two-level networks shows 100% accuracy for non-fall activities and fall events. The data sets for training and testing have been collected using the devices we assembled with four off-the-shelf components. We have programmed them using a prototype of our proposed framework. The data sets include seven types of nonfall activities and four types of fall events. This paper advances the state of the art for the development and training of wearable devices for monitoring non-fall activities and detecting fall events.Index Terms-Fall detection, learning algorithms, neural networks, MEMS, 9-axis sensor, semi-automatic extraction of training examples, softmax regression.

show abstract

“…Prior approaches have either used the internal joints of the robot to resist disturbances (e.g. ankle strategy, hip strategy [1], [2], or inertial shaping [3]) or external contacts (e.g., protective stepping [4], knee contact [5], hand contact [6], [7]). However, these strategies have been considered in isolation and there have been limited attempts to unify multiple strategies into a single fall mitigation system.…”

Section: Introductionmentioning

confidence: 99%

Unified Multi-Contact Fall Mitigation Planning for Humanoids via Contact Transition Tree Optimization

Wang

Hauser

2018

2018 IEEE-RAS 18th International Conference on Humanoid Robots (Humanoids)

View full text Add to dashboard Cite

This paper presents a multi-contact approach to generalized humanoid fall mitigation planning that unifies inertial shaping, protective stepping, and hand contact strategies. The planner optimizes both the contact sequence and the robot state trajectories. A high-level tree search is conducted to iteratively grow a contact transition tree. At each edge of the tree, trajectory optimization is used to calculate robot stabilization trajectories that produce the desired contact transition while minimizing kinetic energy. Also, at each node of the tree, the optimizer attempts to find a self-motion (inertial shaping movement) to eliminate kinetic energy. This paper also presents an efficient and effective method to generate initial seeds to facilitate trajectory optimization. Experiments demonstrate show that our proposed algorithm can generate complex stabilization strategies for a simulated planar robot under varying initial pushes and environment shapes. 158.21 J 163.75 J 68.71 J 12.15 J 26.97 J 60.19 J <0.1 J 55 J 90.18 J 134.20 J 62.78 J 85 J 194. 24 J 210.98 J 8.87 J 45.56 J 16.02 J <0.1 J

show abstract

Direction-changing fall control of humanoid robots: theory and experiments

Cited by 36 publications

References 28 publications

Learning a unified control policy for safe falling

Learning a unified control policy for safe falling

Semi-Automatic Extraction of Training Examples From Sensor Readings for Fall Detection and Posture Monitoring

Unified Multi-Contact Fall Mitigation Planning for Humanoids via Contact Transition Tree Optimization

Contact Info

Product

Resources

About