In dynamic environments shared by humans and robots, mobile manipulators must continuously adapt to real-time changes to execute tasks successfully. While global planning methods are effective at considering the full task scope, they lack the computational efficiency required for reactive adaptation. In contrast, local planning approaches can be executed online but are limited by their inability to account for the full task's duration.
To tackle this, we propose Globally-Guided Geometric Fabrics (G3F), a framework for real-time motion generation along the full task horizon, by interleaving an optimization-based planner with a fast reactive geometric motion planner, called geometric fabrics.
The approach adapts the path and explores alternative target poses, while accounting for collision avoidance and the robot's physical constraints.
This results in a real-time adaptive framework considering whole-body motions, where a robot operates in close proximity to other robots and humans.
We validate our approach through various simulations and real-world experiments on mobile manipulators in multi-agent settings, achieving improved success rates compared to vanilla geometric fabrics, prioritized rollout fabrics and model predictive control.
Index Terms: Path and motion planning, geometric fabrics, optimization-based planning, multi-agent environments
- Unsuccessful grasps -
- Successful grasps -
G3F explores alternative grasp poses based on the current state of the dynamic environment, allowing both robots to reach their target
The proposed approach optimizes the path from start to goal for a pick-and-place task, adapting the grasp and place pose accordingly.
(a)
(b)
In the presence of a human agent, G3F is able to react to the human changing the target pose online by moving the object, while avoiding collisions with the human.
(a)
(b)
In the presence of static obstacles, G3F achieves its goal considering the full task horizon from start to goal.
(a)
(b)
In the presence of static agents, G3F achieves its goal by adapting its grasp and place pose while considering the full task horizon.
In this decentralized multi-agent environment, the robots solve a crossover scenario due to the globally-guided solution via G3F.
The proposed G3F outperforms Geometric Fabrics (GF), Prioritzed Rollout Fabrics (P-RF) and MPC in terms of success rate over all four scenarios consisting of 20 randomized multi-agent environments.
Alternative target poses are explored based on the current state of the robot and the dynamic environment. In the Figure below, it is shown that the target pose is optimized with respect to the current end-effector pose of the robot. The axis along which the rotations are allowed, can vary based on the task.
Parameter | Value | |
---|---|---|
QP | solver-type | OSQP |
εabsolute | 10-3 | |
εrelative | 10-3 | |
max iterations | 1000 | |
H | 5 | |
T | 5 or 10 | |
Collision links | [chassis_link, arm_upper_wrist_link] | |
Collision link radii | [0.65, 0.15] rad | |
RF | K | 500 |
Δt | 0.05 s | |
αvelocity | 0.7 | |
Collision links | [chassis_link, arm_upper_wrist_link] | |
Collision link radii | [0.55, 0.2] rad | |
GF | Collision links | [chassis_link, arm_shoulder_link, arm_forearm_link, arm_lower_wrist_link, arm_upper_wrist_link, arm_end_effector_link] |
Collision link radii | [0.45, 0.1, 0.1, 0.1, 0.1, 0.1] rad |
Safe and stable motion primitives via imitation learning and geometric fabrics
In Robotics: Science and Systems, Workshop on Structural Priors as Inductive Biases for Learning Robot Dynamics,
2024.
Reactive grasp and motion planning for adaptive mobile manipulation among obstacles
In Robotics: Science and Systems, Workshop on Frontiers of Optimization for Robotics,
2024.
Multi-Robot Local Motion Planning Using Dynamic Optimization Fabrics
In Proc. IEEE International Symposium on Multi-Robot and Multi-Agent Systems,
2023.