HELIOS uses a large language model to automate an indirect low-thrust trajectory-optimization workflow. Given a physical problem in natural language, it performs Pontryagin’s Minimum Principle derivation, verifies expressions with SymPy, generates C++ shooting code, and solves the resulting numerical problem. The paper evaluates 11 progressively difficult scenarios, including rendezvous, multi-leg stay transfers, gravity assists, and solar-sail minimum-time transfers. It reports a best compilation success rate of 100% (11/11) and compares eight open-source LLM backends, with total scores ranging from 250 to 905. The proposed architecture adapts to constraints and dynamics models without rewriting the core template.
No heat snapshots are available in the last 24 hours.