Reasonable Motion: ASP-based trajectory method, Argoverse 2 eval

Reasonable Motion, a new answer set programming based hybrid quantitative-qualitative method, computes constrained branching trajectory modes for moving objects and was submitted to arXiv on 24 Jun 2026 (arXiv:2606.25626). The paper, by Julius Monsen, Jakob Suchan, Mehul Bhatt and Lars Karlsson, was accepted at LPNMR 2026, scheduled for 7 - 11 September 2026 in Klagenfurt, Austria.

What is Reasonable Motion?

Reasonable Motion is an ASP foundation that enumerates geometrically admissible motion behaviours as stable models, each representing a distinct trajectory mode. The method produces trajectory modes characterised by domain-dependent and domain-independent factors such as derived event sequence, map topology, and domain norms, and it frames each admissible motion behaviour as a stable model that can be inspected and traced back to its logical justification.

The authors position the approach as hybrid: it combines quantitative geometric constraints with qualitative, logic-based reasoning in an answer set programming (ASP) framework. That combination yields enumerated, verifiable trajectory candidates rather than a single learned prediction.

How does the method work and what was evaluated?

The method performs constrained traversal of an environment graph, using ASP to enumerate geometrically admissible motion behaviours and output them as stable models; those models embody both geometric feasibility and higher-level norms or event sequences. In the paper the authors state the system enumerates "geometrically admissible motion behaviours as stable models," and notes that each stable model constitutes a distinct trajectory mode traceable to its underlying logic.

The paper includes an empirical evaluation using Argoverse 2, described on the arXiv page as a "large-scale real-world autonomous driving benchmark representative of the class of dynamic domains within the scope of the proposed method." The evaluation demonstrates applicability to autonomous driving style scenarios, showing the method can operate on realistic map topology and movement characteristics encountered in such domains.

Why it matters

By producing enumerated trajectory modes tied to explicit stable models, Reasonable Motion offers verifiable interpretability that the authors contrast with purely learned approaches. That matters for domains like autonomous driving where traceability of a chosen trajectory to constraints, event sequences and encoded norms is important for diagnostics, safety analysis and regulatory scrutiny. The paper’s acceptance at LPNMR 2026 and its experiments on Argoverse 2 indicate the approach targets both the logic programming community and applied motion-planning benchmarks.

The hybrid quantitative-qualitative framing also broadens the toolset for movement trajectory computation: rather than replacing learning, it supplies a formally grounded alternative that can be inspected, constrained and enumerated by design.

What to watch

Look for the authors’ presentation at LPNMR 2026, 7 - 11 September 2026 in Klagenfurt, Austria, where methodological details and results from the Argoverse 2 evaluation will be available. Also check for released supplemental materials linked from the paper, such as code or data, which the arXiv entry lists under "Code, Data and Media Associated with this Article."