Long-Horizon Multi-Robot Rearrangement Planning for Construction Assembly (2106.02489v3)

Abstract: Robotic assembly planning enables architects to explicitly account for the assembly process during the design phase, and enables efficient building methods that profit from the robots' different capabilities. Previous work has addressed planning of robot assembly sequences and identifying the feasibility of architectural designs. This paper extends previous work by enabling planning with large, heterogeneous teams of robots. We present a planning system which enables parallelization of complex task and motion planning problems by iteratively solving smaller subproblems. Combining optimization methods to solve for manipulation constraints with a sampling-based bi-directional space-time path planner enables us to plan cooperative multi-robot manipulation with unknown arrival-times. Thus, our solver allows for completing subproblems and tasks with differing timescales and synchronizes them effectively. We demonstrate the approach on multiple case-studies to show the robustness over long planning horizons and scalability to many objects and agents of our algorithm. Finally, we also demonstrate the execution of the computed plans on two robot arms to showcase the feasibility in the real world.

Long-Horizon Multi-Robot Rearrangement Planning for Construction Assembly

This paper presents a sophisticated approach to address the challenges in long-horizon multi-robot rearrangement planning, specifically targeting construction assembly tasks. The authors propose a novel planning system designed to coordinate large, heterogeneous teams of robots, allowing for parallelization and scalability in complex, task-oriented environments.

Overview of the Approach

The paper builds upon previous work in robotic assembly planning by introducing a decomposition-based method for handling large teams of robots and extensive planning horizons. It combines optimization techniques with sampling-based path planning to address manipulation constraints in a cooperative, multi-robot system where arrival times are unknown. This integration allows for synchronizing tasks with varying timescales, enhancing coordination and efficiency across robotic teams.

Key Components

1. Problem Decomposition: The approach decomposes the overall planning problem into smaller subproblems, each consisting of a subset of robots and tasks. This allows for scalability and more manageable computation while ensuring that planned trajectories are feasible within larger cooperations.
2. Time-Embedded Keyframes Optimization: The authors employ an optimization method to solve for manipulation constraints at mode switches between task assignments. By sampling time embeddings across manipulation constraints, the technique generates consistent keyframes that are then used as goals for the path-planning components.
3. Bi-directional Space-Time RRT Planning: The paper introduces a novel bi-directional Rapidly-exploring Random Tree (RRT) algorithm in space-time, which efficiently connects keyframes while accounting for unknown arrival times. This ensures collision-free paths amidst dynamic environments where other robots are already moving according to previously fixed trajectories.

Demonstrations and Results

The paper demonstrates the robustness and scalability of the proposed method through various simulations on complex construction models, including towers, walls, wells, and pavilions with up to 113 parts and 12 heterogeneous agents. Results indicate effective execution times and computation times, showcasing improved utilization of robots without compromising feasibility.

Additionally, the feasibility of computed plans is validated through real-world experiments, highlighting the practicality of the approach in physical construction settings.

Implications and Future Directions

The contributions of this paper lie in its innovative method of decomposing and solving multi-robot planning problems by leveraging time-embedded optimization and bi-directional planning. This provides a significant enhancement to existing construction assembly processes, potentially automating tasks that require delicate coordination among varied robotic systems.

Future research could explore extensions of this framework to accommodate force and torque constraints, adapt to dynamic environmental changes, and incorporate uncertainty in robot actions and task durations. Enhancements in real-time planning capabilities, further reduction of computational overhead, and integration with machine learning approaches for improved prioritization and task assignment indicate promising directions for advancing autonomous construction assembly.

In conclusion, the planning system proposed in this paper represents a comprehensive solution to long-horizon, multi-robot assembly challenges, highlighting a productive intersection of logic-geometric programming and sampling-based methodologies in dynamic environments.