GLAS: Global-to-Local Safe Autonomy Synthesis for Multi-Robot Motion Planning with End-to-End Learning

Benjamin Rivière, Wolfgang Hönig, Yisong Yue, Soon-Jo Chung

Abstract

We present GLAS: Global-to-Local Autonomy Synthesis, a provably-safe, automated distributed policy generation for multi-robot motion planning. Our approach combines the advantage of centralized planning of avoiding local minima with the advantage of decentralized controllers of scalability and distributed computation. In particular, our synthesized policies only require relative state information of nearby neighbors and obstacles, and compute a provably-safe action. Our approach has three major components: i) we generate demonstration trajectories using a global planner and extract local observations from them, ii) we use deep imitation learning to learn a decentralized policy that can run efficiently online, and iii) we introduce a novel differentiable safety module to ensure collision-free operation, thereby allowing for end-to-end policy training. Our numerical experiments demonstrate that our policies have a 20% higher success rate than optimal reciprocal collision avoidance, ORCA, across a wide range of robot and obstacle densities. We demonstrate our method on an aerial swarm, executing the policy on low-end microcontrollers in real-time.

Introduction: The Centralized vs. Decentralized Trade-off

Multi-robot motion planning usually forces a trade-off between centralized and decentralized approaches. Centralized planners are "complete" (they find a solution if one exists) and avoid local minima, but they are computationally expensive and require full knowledge of every robot's state—a luxury rarely available in the real world. Conversely, decentralized controllers (like ORCA or Artificial Potential Fields) are fast and only need local sensor data, but they are notorious for getting "stuck" in local minima (e.g., a U-shaped obstacle) or gridlock.

The GLAS (Global-to-Local Safe Autonomy Synthesis) framework seeks to bridge this gap. It synthesizes a local, decentralized policy that inherits the sophisticated decision-making of a global planner while guaranteeing collision-free operation through a novel differentiable safety module.

The GLAS Framework

The GLAS framework consists of three main pillars designed to move from global intelligence to local execution:

  1. Expert Demonstration & Imitation Learning: The process begins with a centralized, resolution-complete global planner generating thousands of successful trajectories. A neural network is then trained to map local observations—relative positions of nearby robots and obstacles—to the actions the global expert took.
  2. Deep Sets Architecture: To handle a variable number of neighbors and obstacles, GLAS utilizes a "Deep Sets" neural network. This architecture is permutation invariant, meaning the policy's output doesn't change if you swap the order in which neighbors are listed in the input vector. This is significantly more computationally efficient than using CNNs on occupancy grids.
  3. Differentiable Safety Module: This is the paper's core theoretical contribution. Instead of using a standard optimization-based safety filter (which is hard to backpropagate through), the authors derived an analytic, differentiable safety function based on Control Barrier Functions (CBFs).
The GLAS pipeline: from global expert demonstrations to a decentralized, safe policy.

End-to-End Training

Because the safety module is differentiable, the neural network is trained through the safety layer. The policy learns to "smoothly interact" with the safety constraints, leading to lower control effort and better coordination. Training end-to-end through the safety module is superior to training a policy in isolation and applying a safety filter later; when the network "knows" the safety module will intervene, it learns to produce more efficient, less aggressive controls.

Architecture of the GLAS policy, featuring Deep Set modules for permutation invariance and a differentiable safety module.

Experimental Evaluation

The authors generated a massive synthetic dataset for training and validation, consisting of approximately 200,000 demonstrations resulting in 40 million data points. Scenarios included varying numbers of robots (4, 8, and 16) in environments with 10% or 20% obstacle density.

Key Results

  • Success Rate: GLAS achieved a 20% higher success rate than ORCA across various robot and obstacle densities. It specifically excelled in avoiding the local minima that caused ORCA-controlled robots to get stuck.
  • Efficiency: The end-to-end trained model used significantly less control effort (up to 6.25% less) compared to a "two-stage" training approach.
  • Real-Time Hardware: The policy was implemented in C and deployed on Crazyflie 2.0 quadrotors. Evaluation took only 3.4ms to 5.0ms on a 168 MHz STM32 microcontroller, proving its viability for resource-constrained aerial swarms.
  • Generalization: While trained on up to 16 robots, the policy showed the ability to generalize to 32-robot scenarios, though performance naturally degrades as density increases beyond the training distribution.

Conclusion and Future Work

GLAS demonstrates that robots can avoid many global-scale traps using only local observations, provided they have been trained on enough expert global trajectories. By deriving an analytic form for safety rather than solving a Quadratic Program (QP) at every step, the system is fast enough for microcontrollers while remaining mathematically provable.

While GLAS avoids many local minima, it is still a local policy and cannot theoretically guarantee completeness in every possible cluttered environment. Future work includes exploring Adaptive Dataset Aggregation (like DAgger) to better handle out-of-distribution states and investigating Graph Neural Networks (GNNs) for even better relational modeling between robots.

Summary updated

Abstract

We present GLAS: Global-to-Local Autonomy Synthesis, a provably-safe, automated distributed policy generation for multi-robot motion planning. Our approach combines the advantage of centralized planning of avoiding local minima with the advantage of decentralized controllers of scalability and distributed computation. In particular, our synthesized policies only require relative state information of nearby neighbors and obstacles, and compute a provably-safe action. Our approach has three major components: i) we generate demonstration trajectories using a global planner and extract local observations from them, ii) we use deep imitation learning to learn a decentralized policy that can run efficiently online, and iii) we introduce a novel differentiable safety module to ensure collision-free operation, thereby allowing for end-to-end policy training. Our numerical experiments demonstrate that our policies have a 20% higher success rate than optimal reciprocal collision avoidance, ORCA, across a wide range of robot and obstacle densities. We demonstrate our method on an aerial swarm, executing the policy on low-end microcontrollers in real-time.
BibTeX
@article{riviere2020glas,
  author = {Benjamin Rivière and Wolfgang Hönig and Yisong Yue and Soon-Jo Chung},
  title = {{GLAS: Global-to-Local Safe Autonomy Synthesis for Multi-Robot Motion Planning with End-to-End Learning}},
  journal = {IEEE Robotics Autom. Lett.},
  year = {2020},
  url = {https://openreview.net/forum?id=WAllGno2Kb},
}
Figure