Collision detection in legged locomotion using supervised learning

We propose a fast approach for detecting collision- free swing-foot trajectories for legged locomotion over extreme terrains. Instead of simulating the swing trajectories and checking for collisions along them, our approach uses machine learning techniques to predict whether a swing trajectory is collision-free. Using a set of local terrain features, we apply supervised learning to train a classifier to predict collisions. Both in simulation and on a real quadruped platform, our results show that our classifiers can improve the accuracy of collision detection compared to a real-time geometric approach without significantly increasing the computation time.

Authors

• Finale Doshi
• Emma Brunskill
• Alexander C. Shkolnik
• Thomas Kollar
• Khashayar Rohanimanesh
• Russ Tedrake
• Nicholas Roy

Keywords

• Legged locomotion
• Supervised learning
• Leg
• Trajectory
• Foot
• Testing
• Solid modeling
• Robots
• Predictive models
• Kinematics
• Collision Detection
• Machine Learning
• Landforms
• Collision-free Trajectory
• Heuristic
• Decision Tree
• Line Model
• Classifier Training
• Convex Hull
• Geometric Model
• Simple Classification
• Secondary Loss
• Linearly Separable
• Potential Step
• Robot Model
• Position Of The Robot
• Terrain Slope
• Real Robot
• AdaBoost Classifier
• Rapidly-exploring Random Tree
• Rough Terrain
• Cylinder Model
• Real-time Trajectory
• Quadruped Robot
• Terrain Height
• Input Space
• 5-fold Cross-validation
• Central Body