"The first term builds the parts. The second term asks them to survive each other."
A Two-Term Embodied Curriculum
Two-semester sequence gives Teaching with This Book a concrete systems role: use the first term for foundations and the second for integrated systems and capstones. The section keeps asking what the agent observes, what it remembers or updates, which action changes, and what evidence would convince a skeptical reader.
This section develops the technical contract for two-semester sequence into a usable mental model. First we define the object of study, then we connect it to the agent loop, then we test it with a compact implementation.
The key question in Two-semester sequence is practical: what must the agent know, what can it observe, what action is available, and what evidence shows that the action worked under the stated conditions?
Two-semester sequence should be judged by the action it improves. A section claim is strong when it names the decision, the measurement, and the failure mode before a larger model or simulator is introduced.
Theory
For Two-semester sequence, the practical design rule is to make the interface inspectable before optimization begins: inputs, outputs, units, latency, bounds, and failure labels should all be visible in the saved artifact.
The mechanism in Two-semester sequence is the contract between representation and action. Name what enters the module, what leaves it, which assumptions make that transformation valid, and which log would reveal a bad handoff.
Worked Example
For Two-semester sequence, keep one concrete rollout in view. A sensor reading becomes an estimate, the estimate constrains an action, the action changes the world, and the next observation confirms or contradicts the assumption. The section's idea is useful only if it improves that loop.
For Two-semester sequence, the small contract exists to expose the teaching artifact before tooling takes over. Use notebooks, simulators, shared logs, rubrics, and capstone studios only when they preserve the same observation, action, metric, and failure fields.
Practical Recipe
- Write the observation, action, and success metric before choosing a model.
- Build a baseline that is simple enough to debug by inspection.
- Add the library implementation only after the baseline behavior is understood.
- Record failures as structured cases: perception error, state error, planning error, control error, or evaluation error.
- Run at least one perturbation test before trusting the result.
The common mistake in Two-semester sequence is to trust a component score before checking the closed-loop interface. The failure usually appears where state, timing, authority, or evaluation context crosses a module boundary.
A team using Two-semester sequence starts by writing the task panel, not by picking the largest model. They keep a baseline run, a maintained-tool run, and a perturbation run in the same result folder. The comparison is accepted only when the action trace, metric, and failure labels come from one script.
When two-semester sequence feels abstract, ask what would be different in the next frame of video, the next robot state, or the next safety margin.
For Two-semester sequence, the open research question is not whether a larger policy can produce a better demo. The sharper question is whether the method improves reliability across new scenes, new embodiments, delayed feedback, and rare failures under an evaluation protocol that another lab can reproduce.
For Two-semester sequence, can you name the observation, action, protected assumption, success metric, and one likely failure case? If any field is vague, rewrite the contract before adding model complexity.
Topic-Native Deepening
A two-semester sequence is where the book can fully breathe. The first term can establish the physics, estimation, simulation, and policy-learning foundations; the second can move into language, world models, safety, and longer capstones without compressing everything into one overloaded arc.
The course-design challenge is coherence across the handoff. Students should feel that the second term extends the same loop rather than starting a second disconnected subject.
Two-semester sequence becomes teachable once the student can state the operative variables, the decision boundary, and the evidence artifact. The section should therefore be read together with Part V and Part VIII, where the same loop is developed from adjacent angles.
Let semester one build foundation set $F$ and semester two build extension set $E$. The sequence works when prerequisite edges form a sparse DAG from $F$ to $E$, not a tangled graph that forces constant review of forgotten assumptions.
This is why the first semester should overinvest in frames, interfaces, data cards, and evaluation discipline. Those concepts quietly support everything interesting that happens later.
- Use term one for perception, dynamics, control, state estimation, simulation, RL, and imitation.
- End term one with a modest integrative project that proves students can close the loop.
- Open term two with a brief refresh, then add language, VLAs, planning, 3D representation, and deployment topics.
- Run a larger second-term capstone that can draw from both terms without re-teaching prerequisites.
- Keep one shared evidence schema across both semesters so artifacts remain comparable.
| Dimension | What To Specify | Why It Matters |
|---|---|---|
| Term one | Foundations, control, state estimation, simulation, RL, imitation | Technical floor and first integrative project. |
| Interterm artifact | Baseline system plus replay and postmortem | Prevents term-two amnesia. |
| Term two | Language, planning, world models, safety, deployment, frontier topics | Advanced synthesis. |
| Final deliverable | Research-grade capstone with proposal and defense | Uses both halves of the sequence. |
def validate_sequence(payload: dict[str, object]) -> dict[str, object]:
assert payload, "payload must not be empty"
return payload
# Two-term sequence card.
sequence = {
"term_one_project": "simulator-based mobile manipulation baseline",
"term_two_project": "language-conditioned embodied capstone",
"shared_evidence_schema": True,
}
print(validate_sequence(sequence)){'term_one_project': 'simulator-based mobile manipulation baseline', 'term_two_project': 'language-conditioned embodied capstone', 'shared_evidence_schema': True}The expected output should reveal continuity across terms. If the evidence schema changes between semesters, students will struggle to connect the advanced work back to the foundations.
After the from-scratch contract is clear, the practical route uses Same book stack plus course project repositories, CI, shared data cards, simulator presets. The payoff is that standard interfaces, logging, batching, and replay support move from ad hoc glue code into maintained infrastructure, while the evidence schema stays the same.
The strongest two-term designs keep term-one artifacts alive as baselines for term two. That makes progress legible and reduces the temptation to discard hard-won infrastructure every semester.
A frontier teaching opportunity is to let term-two students reproduce or stress-test a current research claim using the infrastructure they built in term one. That is how the sequence becomes a research pipeline instead of two classes.
For Two-semester sequence, the artifact should show the course-design decision, the evidence students must produce, and the failure mode that would trigger a revised assignment or rubric.
- Two-semester sequence matters when it changes an embodied agent's action under a stated observation and metric.
- Use the first term for foundations and the second for integrated systems and capstones.
- Strong evidence is saved as one artifact containing the baseline, the maintained-tool path, the metric panel, and labeled failures.
Design a method-matched experiment for Two-semester sequence. Specify the environment, observation schema, action interface, metric, and one perturbation that targets the section's core assumption.
Section References
Biggs, J. Teaching for Quality Learning at University. Open University Press, 1999.
Use for constructive alignment between learning outcomes, activities, and assessment.
Anderson, L. W. and Krathwohl, D. R. A Taxonomy for Learning, Teaching, and Assessing. Longman, 2001.
Use for designing assessments that move from recall to analysis, creation, and evaluation.
What's Next?
Next, continue with the following teaching section, where the Two-semester sequence contract becomes a concrete course-design decision.