Reference

Glossary

Every term used across the programme, the simulator interface and the mission debrief, grouped by domain and searchable.

04Terminology

Definitions in use

Where a term carries a specific operational meaning in this programme, that meaning is given here.

30 terms

Core model

Algorithmic confidence: The agent's own estimate of how reliable its perception and planning currently are. High confidence pushes the operating point toward autonomy.
HAPM: The Human-AI Partnership Model: a framework for allocating tasks between operator and autonomous agent as interdependent team members along a dynamic spectrum.
Operating point: The current position along the Spectrum of Shared Control for a given task.
Shared autonomy: A control paradigm in which authority over a task is distributed continuously between human and machine rather than assigned to one of them.
System-Level Operator: An operator whose job is to manage a system of platforms and autonomy levels toward a mission outcome, rather than to pilot one vehicle.

Control

Arbitration: The mechanism that blends operator input and agent intent into a single command stream, deciding moment by moment whose intent dominates.
Direct Manual Control: Mode 01. The operator performs kinematic execution, path planning and objective formulation; the machine holds only passive safety envelopes.
Dynamic Arbitration: Mode 03. The operator supervises high-level goals and approves tactics; the agent estimates intent and synthesises trajectories.
Handover: A transition of authority between operator and agent. Safe handovers are announced, justified and preserve the state the receiving party needs.
Shared Assistance: Mode 02. The operator continuously guides the trajectory while the machine smooths kinematics, avoids obstacles and supplies force feedback.
Supervisory Autonomy: Mode 04. The operator sets strategic objectives and monitors performance while the agent executes the mission plan and coordinates agents.
Teleoperation: Direct remote control of a platform, with the operator closing the perception-action loop over a communications link.
Waypoint: A commanded position the platform navigates to autonomously, used when the operator directs outcomes rather than motion.

Human factors

Agency alienation: The state in which an operator no longer experiences themselves as an actor in the system and becomes a passive observer, reducing engagement and slowing response to handovers.
Agency dilution: The weakening of an operator's felt authorship of an outcome when assistive automation silently modifies their input.
Cognitive load: The share of the operator's attention already committed. Rising load is a trigger to shift authority toward the agent.
Sense of agency: The operator's subjective experience of causing and owning an outcome — a design objective in its own right, not a side effect.
Trust calibration: Aligning how much the operator relies on the agent with how reliable it actually is in the current conditions.

Sensing

LiDAR: Laser ranging used to build precise 3D geometry of terrain, vegetation and disturbance patterns, independent of light conditions.
Modality: A class of sensing — thermal, LiDAR, RGB, macro imaging, chemical — each with its own strengths, footprint shape and failure conditions.
SLAM: Simultaneous Localisation and Mapping: building a map of an unknown environment while tracking the platform's own position within it.
Thermal imaging: Infrared sensing that reveals heat signatures, useful for recent biological activity and for working after dark.

Forensics

Chain of custody: The documented, unbroken record of who handled a piece of evidence, when, and how — required for the evidence to be admissible.
Contamination: Any alteration of the scene or a trace caused by the recovery operation itself, including tracks, downwash, heat and physical contact.
Evidence integrity: The degree to which a recovered trace remains in the state it was found, physically and evidentially.
Trace: Any physical remnant of the crime that can be recovered and analysed — tissue, fluid, tool marks, tracks, discarded equipment.

Simulator

Detection ping: The visual and audible signal raised when a platform's sensor footprint intersects an undiscovered evidence trace.
Footprint: The ground area currently covered by a platform's active sensor, drawn per modality in the colourblind-safe palette.
Full team deployment: Running the scene with all five platforms at once, so the operator coordinates a heterogeneous fleet rather than a single robot.
Heterogeneous fleet: A team of platforms with different mobility and sensing profiles, deployed so their capabilities complement each other on one scene.