Topic 1 — Capstone Overview & Project Brief
The capstone is not “just another lab.” It is a single, coherent project that integrates the concepts, tools, and stacks you have built across Chapters 1–6 into a demonstrable Physical AI system. This topic helps you define a clear project brief so that every design decision—from architecture to evaluation—traces back to explicit goals and constraints.
1.1 Defining Your Problem Space
Start by answering three questions:
- What is the core capability you want to demonstrate?
Examples:- Language-driven pick‑and‑place in a cluttered scene.
- Humanoid inspection of a simulated lab or warehouse.
- Multi-room patrol with natural-language updates to a human operator.
- Where does it live in the stack?
- Does it stress perception more than planning?
- Is autonomy or multi-agent coordination central to the story?
- Is the focus on reliable task execution, human interaction, or both?
- What is realistically achievable in 4–6 weeks?
- Scope down to a narrow but end-to-end experience rather than many partial features.
Write these into a one-page problem statement that you can share with peers or mentors for feedback.
1.2 Project Constraints, Assumptions & Non‑Goals
Great capstones are as much about what you don’t build as what you do.
- Constraints:
- Compute (e.g., single GPU, simulation-only, limited memory).
- Time (weeks remaining, course schedule).
- Hardware (simulation-only vs. optional physical robot access).
- Assumptions:
- Indoor, structured environments vs unstructured outdoor environments.
- Fixed number of robots (one humanoid, plus optional assistants).
- Network reliability (local host vs distributed machines).
- Non-goals (intentionally out of scope):
- Full human-safe physical deployment.
- Production-grade cloud orchestration.
- General-purpose open-world behavior.
Documenting non-goals lets you avoid feature creep and focus on depth.
1.3 Success Metrics & Evaluation Criteria
Define objective measures of success before writing code:
- Task metrics:
- Task completion rate (e.g., successful deliveries, inspections, or pick‑and‑place episodes).
- Time to complete a mission or scenario.
- Number of recovery behaviors triggered (replans, retries).
- Quality metrics:
- Navigation smoothness (collisions, oscillations, deadlocks).
- Perception accuracy where relevant (detection/segmentation success for key objects).
- Usability of the human interface (clarity of logs, instructions, or natural-language outputs).
- Robustness metrics:
- Performance under small perturbations (noisy sensors, blocked paths).
- Ability to resume after partial failures (node restart, robot dropout in multi-agent scenarios).
These metrics will later inform your experiments and demo script.
1.4 Deliverables Checklist
By the end of this topic, you should have:
- A written project brief (1–2 pages) that includes:
- Problem statement and project title.
- Target capabilities and user story (“What can the robot do?”).
- Constraints, assumptions, and explicit non-goals.
- Preliminary success metrics and evaluation criteria.
- A rough sketch of environments, robots, and interaction modes (e.g., single humanoid in an apartment map; two robots in a small warehouse).
1.5 Mini-Lab: Draft and Review Your Brief
Goal: Produce a first-draft capstone brief and pressure-test it.
Tasks
- Write a 1–2 page brief following Sections 1.1–1.4.
- Share it with at least one peer, mentor, or teaching staff member:
- Ask: “Is this achievable in 4–6 weeks with my current skills?”
- Ask: “Where is the biggest technical risk?”
- Revise the brief based on feedback and lock it in as your capstone contract.
Deliverables
- Draft and final versions of your project brief (kept in version control).
- A short review note summarizing feedback and changes you made.
Summary
A strong capstone starts with a well-scoped brief: a clear problem statement, explicit constraints, and measurable success criteria. In the next topic, you will turn this brief into a concrete system architecture and design blueprint that guides implementation.