Core Concepts
These concepts define how entities and data flow work in Ticos. Understanding them makes configuration and troubleshooting much easier.
Robot
A Robot is the core managed entity: one physical device. Each robot has:
- Unique identifier — For communication, data association, and scoping |
- Robot model — Hardware type; defines 3D model and sensor config |
- Agent — The currently bound “AI brain” config |
- Health and telemetry — Live hardware and software metrics |
Think of a robot as body + agent: the body is the hardware and sensors; the agent is dialogue, decisions, and tools.
Agent
The Agent is the robot’s “brain”—how it understands input, decides what to do, and interacts with people. An agent configuration includes:
| Item | Purpose |
|---|---|
| Model | LLM choice (e.g. GPT-4, Claude, DeepSeek) and parameters |
| Personality | System prompt: role, tone, and behavior rules |
| Voice | TTS voice and speed |
| Expressions | Mapping from emotion to facial expressions |
| Actions | Executable physical actions (wave, nod, etc.) |
| Skills | MCP tools and capability modules |
| Knowledge | RAG knowledge base for domain expertise |
The same agent can be bound to multiple robots so you can reuse one “persona” across the fleet.
Scene
A Scene describes the environment and navigation context where the robot operates:
- Map — 2D/3D map data |
- Routes — Predefined paths and waypoints
- Zones — Functional areas (e.g. charging, restricted)
- Task nodes — Key locations for actions
Scenes can be shared across robots so you configure once and reuse.
Teleoperation (Teleops)
Teleops is the live control center for monitoring and manual control:
- Video — Low-latency WebRTC stream (media service) |
- Sensor panels — LiDAR, IMU, depth, etc. |
- Control — Keyboard/gamepad input turned into motion commands |
Teleops uses a draggable grid layout: you can rearrange and resize panels and save presets.
Robot Model
A Robot model is a hardware template maintained by admins under “Robot models”. It defines:
- URDF — 3D model for visualization
- Sensors — Supported sensor list for this model
- Defaults — Default motion and safety parameters
When you register a new robot you pick a model; that determines which sensors and visualizations are available.
Communication
Ticos uses two layers:
| Layer | Service | Use |
|---|---|---|
| Telemetry | Real-time data channel (REST API) | Sensors, status, control |
| Media | Media service (WebRTC) | Video and audio |
The data channel is pub/sub for high-frequency telemetry; the media service handles encrypted video and voice.
MCP Tools
MCP (Model Context Protocol) tools extend what the agent can do, for example:
- Hardware drivers (arm, base controller, etc.) |
- Perception APIs (person detection, object recognition) |
- External services (weather, DB, business APIs) |
Tools register via a standard protocol; the agent discovers and calls them in conversation without hard-coded glue.