The five factors are the structural core of ANIMA's self-assessment. Each evaluates at Pre / Runtime / Post, and the composite yields a success probability with calibrated confidence. GOA composes multiplicatively, SQA learns from history, and PEA makes the system better over time.
Intent-to-semantic alignment. Upstream drift is penalized directly: MQA_ita = 1 − drift_score. L0 surfaces drift; ITA carries it into the final probability.
Trajectory smoothness, contact force envelope, grasp margin — sub-indicators read from simulation or real-hardware sensors. soma-care v0.4 is expanding this factor.
Beta prior updated from a rolling success-rate read of `pea_log.jsonl`. Starting with v0.2, p_skill stopped being a constant.
Goal success probability, composed multiplicatively: P(success) = ∏ Pᵢ. Averaging would mask low-probability bottlenecks; multiplication makes any single weak link instantly visible.
Preference-experience retrieval, weighted by three factors: recency × 0.5 + relevance × 3.0 + importance × 2.0. Each task appends one row to `pea_log.jsonl` that future planning retrieves.
The most concrete reference implementation today
A tabletop language-driven robot arm project used to validate ANIMA's cognition loop, visual grounding, skill execution, and failure recovery.
It is not the final product. It is the benchmark and manipulation capability layer on the path to home robotics.
Medical-care intent-to-action simulation
The medical-care product line of the SOMA family. Powered by the ANIMA cognition framework, it turns intent (text, voice, and future BCI signals) into auditable robot actions inside a simulated hospital ward. v0.4 focuses on showcasing the full intent-to-care loop via MuJoCo simulation and offline video.
The care product line of the SOMA family, running in parallel with soma-arm and sharing the ANIMA cognition framework.