Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Motion Planning Under Uncertainty with Complex Agents and Environments via Hybrid Search
Authors: Daniel Strawser, Brian Williams
JAIR 2022 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | A variety of 2D and 3D test cases are presented including a linear case, a Dubins car model, and an underwater autonomous vehicle. The method is shown to outperform other methods in terms of speed and utility of the solution. Additionally, the models of trajectory risk are shown to better approximate risk in simulation. Section 10: Results, Four sets of test cases are included. Benchmarks were performed on a machine with an Intel i7-4790 CPU at 3.60 GHz with 16 GB RAM. GPU tests utilized an Nvidia Ge Force 1080 Ti. |
| Researcher Affiliation | Academia | Daniel Strawser EMAIL Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge, MA 02139 Brian Williams EMAIL Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge, MA 02139 |
| Pseudocode | Yes | Algorithm 1: First Feasible Hybrid Search. Algorithm 2: Path Generator. Algorithm 3: Generate Next Best Child. Algorithm 4: Avoid Path Loop. Algorithm 5: Expand Neighbors. Algorithm 6: Rewire Path. Algorithm 7: Validate Trajectory. Algorithm 8: Shooting Method Monte Carlo. Algorithm 9: Collocation Simulation. Algorithm 10: Shooting Method Simulation. |
| Open Source Code | No | The paper does not provide any explicit statement or link regarding the availability of open-source code for the described methodology. |
| Open Datasets | No | The paper describes various 'test cases' and 'scenarios' such as an 'AUV exploring the undersea volcano Kolumbo' or a 'Dubins car model' but does not provide concrete access information (links, DOIs, formal citations) for any publicly available or open datasets used for these experiments. The environments and models are described, but not as formally accessible datasets. |
| Dataset Splits | No | The paper focuses on motion planning and simulation with test cases, rather than traditional machine learning experiments that involve splitting a dataset into training, testing, and validation sets. Therefore, no dataset split information is provided. |
| Hardware Specification | Yes | Benchmarks were performed on a machine with an Intel i7-4790 CPU at 3.60 GHz with 16 GB RAM. GPU tests utilized an Nvidia Ge Force 1080 Ti. |
| Software Dependencies | Yes | The algorithms are implemented in C++ and, for GPU-based parallelization, Nvidia's Cuda. SNOPT Version 7.6 was used as an optimizer for the trajectory planner (Gill et al., 2017). |
| Experiment Setup | Yes | For each test, the risk was set to ϵ = 0.20, e.g. a 20% probability of failure. Minimizing the trajectory's Euclidean distance was used for an objective. For this study, ct was set to 100, cinflection set to 10,000 to account for the approximately 10k J used for a change in the buoyancy, and cyaw set to 2. For this example, µ = [-5.0, 0, -5.0, 0, 0, 0] and σx = 5.0, σy = 0, σz = 5.0, σψ = 0, σθ = 0, σφ = 0 (once again, all off-diagonal terms in the covariance matrix were zero). For these benchmarks, the chance constraint, ϵ, was set to 0.15. For the two sampling-based methods, α was set to 5e6 and the number of scenarios S set to 2500. Uncertainty in the agent's dynamics, represented by σx, σy, σθ in Eqs. (97)–(99) was set to 0.1 for all variables. |