Publication record · 18.cifr/2014.schulman.trajopt

Motion planning with sequential convex optimization and convex collision checking

v1.0.0

John Schulman (University of California at Berkeley), Yan Duan (University of California at Berkeley), Jonathan Ho (University of California at Berkeley), Alex Lee (University of California at Berkeley), Pieter Abbeel (University of California at Berkeley)

RAI18.cifr/2014.schulman.trajopt

The International Journal of Robotics Research· 2014· doi:10.1177/0278364914528132

We present a new optimization-based approach for robotic motion planning among obstacles. Our algorithm uses sequential convex optimization with hinge-loss collision penalties and continuous-time safety constraints. Implemented as TrajOpt, it outperforms CHOMP and OMPL planners on 7-34 DOF robots in planning time and path quality.

motion planningtrajectory optimizationsequential convex optimizationcollision avoidancerobot arms

✦ Research context

What this agent contributes to the literature.

Problem solved

Prior trajectory optimizers like CHOMP relied on potential fields prone to local minima and lacked formal constraint handling, while sampling-based planners were slow and produced low-quality paths. TrajOpt provides fast, reliable collision-free trajectory optimization for robots with many DOF in cluttered environments, starting from straight-line initializations.

Novelty

TrajOpt introduces sequential convex optimization with continuous-time collision checking via swept-volume approximations and signed distance hinge-loss penalties. The penalty-augmented outer loop adaptively increases collision costs until feasibility is achieved, unlike CHOMP's fixed gradient-field formulation. This enables reliable collision-free planning from naïve initializations for high-DOF robots.

Related research

Computing related research...

Canvas contract1-in / 1-out · unpacked into planning_problem, solver_params legacy ports

Sample data

Loading sample data...

Total calls

This month

Citations

Last called

—

Image digest

sha256:ef6c4902cabb554b978efad3bb4de48c81f4449a92513b0ed2e4e06171dc34df

Invoke command

python main.py

Inputs

input:application/json

Outputs

output:application/json

Citation

Loading DOI…

Invoke

CPU compute only

How to get GPU access: Your university, lab, or company can become a CIFR institutional member. Members get GPU-accelerated runs for all their researchers. Contact us

Pre-filled with the paper's canonical scenario. Click Invoke agent to reproduce the original result, or edit the JSON below to run a counterfactual.

inputapplication/jsonoptional

Unified canvas input with planning problem specification

Leave empty to run the paper's canonical scenario.

{
  "planning_problem": {
    "n_dof": 7,
    "n_steps": 10,
    "start_config": [
      0,
      0.5,
      0,
      -1,
      0,
      1.5,
      0
    ],
    "end_config": [
      1.2,
      -0.5,
      0.8,
      -0.5,
      -0.8,
      1,
      0.5
    ],
    "joint_limits": [
      [
        -3.14,
        3.14
      ],
      [
        -2,
        2
      ],
      [
        -3.14,
        3.14
      ],
      [
        -2.5,
        0
      ],
      [
        -3.14,
        3.14
      ],
      [
        0,
        3.14
      ],
      [
        -3.14,
        3.14
      ]
    ],
    "obstacles": [
      {
        "center": [
          0.4,
          0.1,
          0.5
        ],
        "radius": 0.12
      },
      {
        "center": [
          0.3,
          -0.2,
          0.3
        ],
        "radius": 0.1
      }
    ]
  },
  "solver_params": {
    "max_outer_iter": 15,
    "max_inner_iter": 20,
    "collision_penalty_init": 10,
    "penalty_scale_factor": 10,
    "trust_region_init": 0.1,
    "trust_expand": 1.5,
    "trust_shrink": 0.5,
    "trust_min": 0.0001,
    "convergence_tol": 0.001,
    "safety_margin": 0.05
  }
}

Recent invocations(0)

No invocations yet — be the first to call this agent.