Abstract: Sampling-based motion planning is often performed in configuration space, making it a computationally expensive task for manipulators with high degrees of freedom. Proving infeasibility in motion planning, an integrated process of motion planning itself, has the same limitation. Many real-world planning tasks come with constraints for the manipulator, e.g., welding in an exact line, or turning a valve only about its axis, and the constraints are typically represented as manifolds in the configuration space. We present a new approach to sampling-based motion planning with workspace constraints in which we construct an explicit workspace manifold based on the constraints and use a parametric state space in sampling with reduced dimension compared to the configuration space. The parameters include manifold coordinates that map to a unique pose and extra parameters as needed for inverse kinematics calculations. Our method works for manipulators with analytical inverse kinematics solutions, and we apply to the specific anthropomorphic 7-DOF (3R-1R-3R) arm in our experiments. With the reduced dimension of the state space in sampling, we can generate motion plans or infeasibility proofs in less time.

Abstract: Drones and Unmanned Ground Vehicles (UGVs) can be paired together to form symbiotic teams, where drones can quickly move over rough terrain while UGVs can charge and ferry around drones. In this talk, I will introduce two algorithms for planning patrolling paths for drone-UGV teams over indefinite time horizons. These algorithms utilize a second-order cone program that greatly improves performance over classic divide-and-conquer approaches. The results of my numerical simulations and field experiments demonstrate trade-offs in these algorithms and motivate areas of ongoing work.

 

Abstract: Exploring extreme terrain is pertinent for extraplanetary surface exploration and search and rescue missions here on Earth. These high-risk operations are best carried out by autonomous robots, minimizing harm to humans. In this talk, I will describe the difficulties associated with robots traversing extreme terrain and propose an autonomy architecture. The robot follows an encoded procedure of objective synthesis, path planning, adaptive dynamics modeling and control policy generation. I will expand upon each of the procedural steps in the robot’s autonomous exploration algorithm and tell this robot’s story in the context of lunar surface mission seeking ice.