Body Awareness

Proprioception: Knowing Where You Are

Maxim's MovementTracker continuously monitors the robot's kinematic state:

• Angular velocity - Combined yaw + pitch rate (deg/sec)
• Translation velocity - Combined x + y + z movement (mm/sec)
• Angular acceleration - Rate of velocity change (deg/sec²)
• Direction reversals - Detecting thrashing patterns
• Position history - Sliding window for trend analysis

This isn't just logging. The data flows into pain detection, motor learning, and decision-making systems. The robot feels how it's moving.

Pain: Learning from Discomfort

The PainDetector identifies five types of aversive experiences:

Cognitive Pain: Tool Errors

Beyond physical discomfort, Maxim experiences cognitive pain when tools fail. Tool errors are routed through the same PainDetector → NAc → FearAgent pipeline as movement pain, enabling learned aversion to unreliable tools.

Anticipated Pain: Predicting Before Acting

Beyond physical and cognitive consequence-pain, Maxim assesses anticipated pain before an action runs. The PerceivedPainAssessor asks: "if I did this right now, how likely is pain?" and fires a PainType.ANTICIPATED signal whose intensity is max(learned_from_NAc, innate_prior). That felt signal flows through PainBus → hippocampus → the AUT's next LLM context, so the agent can reason about a gut-feeling instead of relying purely on prompt-level logic.

The loop: Layer 2 (PainInterceptorExecutor) fires after a tool touches a sensitive path — the ground-truth signal that trains NAc. Layer 1 (PerceivedPainAssessor) fires before execute, combining that learned knowledge with innate priors. The more sensitive actions the AUT has taken, the more accurate its anticipation becomes — conditioned aversion through experience.

Each pain signal includes:

Pain → Learning

Pain signals don't just trigger immediate responses. They feed into the Nucleus Accumbens through the PainCircuitBridge, creating lasting aversive associations:

Movement Command
       │
       ▼
┌──────────────────┐
│ MovementTracker  │
│ (position data)  │
└────────┬─────────┘
         │
         ▼
┌──────────────────┐
│  PainDetector    │
│ (pattern match)  │
└────────┬─────────┘
         │ PainSignal
         ▼
┌──────────────────┐
│ PainCircuitBridge│
│ (format for NAc) │
└────────┬─────────┘
         │
         ▼
┌──────────────────┐
│ Nucleus Accumbens│
│ (causal learning)│
└──────────────────┘
         │
         ▼
Future: Predict pain BEFORE action

After experiencing pain from a specific action pattern, the NAc learns to predict it. Next time a similar action is proposed, the robot can refuse or modify the plan before experiencing pain again.

Pain Reactions & The SEM Learning Loop

Pain signals don't stay on the PainBus alone. Every PainSignal is also converted into a typed Reaction and dispatched through the ReactionBus. This dual-bus architecture means pain reaches two audiences through two different contracts:

distribute_reward: Pain to NAc Reward Bias

When a pain reaction reaches the NAc via distribute_reward, it adjusts per-node reward_bias values in the Hebbian graph. Negative valence from pain loosens EC similarity thresholds — the system casts a wider net to detect potentially dangerous situations, making it more sensitive to anything that resembles a past painful context.

Pain Spike Episode Boundaries

High-intensity pain (≥ 0.7) triggers the salience_spike_rule in the episode capture pipeline. This forces the current episode to close immediately, capturing the accumulated negative valence, and starts a fresh episode. The effect mirrors how biological trauma creates sharp memory boundaries — the moment of pain becomes the dividing line between "before" and "after" in memory.

Motor Learning: The FocusLearner

Maxim's FocusLearner implements this for gaze control. The problem: camera latency, mechanical dynamics, and tracking delays mean the robot often overshoots or undershoots when following a target.

The solution: Rescorla-Wagner learning to adapt movement gain:

• α = learning rate (0.2 default)
• λ = optimal gain inferred from this trial
• V = current gain estimate

The process:

1. Command movement to target with current gain
2. Observe actual vs expected position (tracking error)
3. Compute optimal gain from overshoot ratio
4. Update gain incrementally toward optimal

Key properties:

• Bounded: Gains constrained to [0.4, 1.0] preventing instability
• Asymptotic: Smooth convergence, no oscillation
• Persistent: Learned gains saved across sessions
• Adaptive: Naturally handles changing conditions

Workspace Bounds Learning

Similarly, the WorkspaceBoundsLearner discovers the robot's reachable space through exploration. Rather than hard-coding limits, the system learns where it can and cannot move, adapting to mounting position, obstructions, and mechanical wear.

Integration: The Embodied Loop

All proprioceptive systems connect to the decision-making loop:

• Movement planning uses learned gains and bounds
• Action proposals are checked against pain predictions
• Execution monitoring detects pain in real-time
• Memory formation records painful episodes for future avoidance
• Energy tracking considers motor effort as a cost

The result: a robot that moves smoothly, avoids harmful patterns, and improves with experience. Not because we programmed every case, but because we gave it the machinery to learn.