A trajectory itself communicates the goal
Aliases: motion legibility · trajectory legibility · goal inference
What it is
Goal legibility from motion is the extent to which early direction, curvature, speed, and posture allow an observer to infer which goal a robot is pursuing. It differs from eventual success: a dynamically correct path can remain equally compatible with several goals until its final segment.
Why it happens
Observers compare partial motion with movements expected under candidate goals and update their belief over time. An early diagnostic deviation raises one goal's posterior probability; motion along a direction shared by several shortest paths adds little information. Body orientation and end-effector posture also enter the inference, so identical base paths need not express identical intent.
Studying it
Researchers can present several candidate goals, truncate trajectories frame by frame, and obtain accuracy-over-time curves. Joint work adds waiting, premature action, and conflict measures. Goal salience, distance, and visibility must be balanced or the scene rather than motion supplies the answer. Real-speed embodied replication captures approach and viewpoint effects absent from video.
Where it stops holding
Exaggeration has costs: a large deviation can increase time, energy, or contact risk. Laboratory benefit may not transfer when observers do not know the candidate set. Dynamic obstacle avoidance can resemble an intentional cue. High-consequence actions still need explicit preview rather than reliance on inferred motion alone.
Applying it
- Check whether the first half of a path differentiates nearby candidate goals; use early curvature or end-effector orientation when necessary.
- Preserve path constraints and safety margin instead of adding theatrical movement through a less safe region.
- Ask untrained observers to predict goals from partial trajectories, measuring first stable correctness, premature action, and total task cost.