Topic 1: Simulation vs Reality — The Deployment Gap

Topic 1 examines why robots that perform reliably in simulation or in tightly controlled lab demos often behave very differently once deployed into real environments. You will connect concepts from Chapter 3 (digital twins and physics engines) with the practical realities of floor surfaces, clutter, human traffic, and long-term wear.

1.1 Physics vs Real Materials

Friction, Slippage, and Surface Variability

Physics engines approximate friction and contact using simple models (e.g., Coulomb friction with a single coefficient). Real floors are messier:

• Polished concrete vs rough concrete vs epoxy coatings
• Dust, spills, tape, cables, and small debris
• Thresholds, ramps, and minor level differences between rooms

For legged robots, this means:

• Small changes in friction can destabilize a gait that looked perfectly stable in Gazebo or Isaac Sim.
• Toe/heel contact points can slip unexpectedly when the robot transitions weight.
• Edge cases (stepping near a cable, painted lines, or a ramp) cause transient instability.

For wheeled bases:

• Turning in place may require more torque than anticipated due to tire scrub.
• Local variations in traction (e.g., near doorways) can cause path-tracking errors.

You will learn to:

• Choose and tune friction coefficients in simulation that resemble your target floor types.
• Design gaits and controllers that are robust across a range of friction values, not a single ideal value.
• Use small-scale physical tests (e.g., friction “patches”) to validate assumptions.

Battery Discharge and Temperature Impact

Simulations often assume a constant supply voltage and ignore temperature. In reality:

• Voltage drops as batteries discharge, reducing peak torque and speed.
• Cold environments reduce effective capacity and current delivery.
• Heat from motors and electronics elevates local temperature, affecting lifetime and performance.

You will:

• Interpret discharge curves and understand how they affect mission duration.
• Incorporate simple battery models into your planning (remaining runtime, derating).
• Learn to include temperature monitoring as part of deployment readiness.

Wear and Tear Over Time

Perfect, rigid links with fixed parameters are a simulation convenience. In deployed systems:

• Gearboxes develop backlash and noise.
• Cables and belts stretch, changing effective kinematics.
• Encoders can accumulate misalignment or dead zones.

Topic 1 introduces:

• How to monitor for long-term drift (e.g., joint offsets, increased current draw).
• Simple calibration procedures to bring a worn robot back into spec.

1.2 Reality-Aware Robotics

Perception Noise and Limited Visibility

Chapter 4 focused on perception algorithms and sensor fusion; Topic 1 revisits them from a deployment angle:

• Sensors accumulate dust, smears, and scratches.
• Lighting in warehouses and factories is inconsistent and can produce glare.
• People, pallets, and carts create frequent occlusions.

You will:

• Design perception stacks that degrade gracefully when certain sensors are partially blocked.
• Use health checks (e.g., histogram analysis of images, laser return rates) to detect sensor degradation.
• Plan periodic cleaning and inspection as part of operational procedures.

Unexpected Obstacles and Humans in the Loop

In the field:

• A single mis-placed pallet can block an entire route.
• Humans may stop to talk near the robot, follow it, or stand in front of it.
• Temporary construction or reconfiguration of shelves changes the environment.

You will study:

• How to design navigation behaviors that assume frequent re-planning, not a static map.
• Human-centered behaviors such as slowing down in dense crowds, yielding in narrow aisles, and avoiding “startling” motions.

Adversarial or Chaotic Environments

Industrial robots must also handle:

• Sensors partially covered by tape, stickers, or accidental damage.
• Network segments going up and down due to maintenance.
• Other robots or forklifts creating complex traffic patterns.

You will learn to:

• Treat these conditions as expected test cases, not rare outliers.
• Build checklists of failure modes that are deliberately exercised in simulation and in safe field trials.

1.3 Closing the Gap

Domain Randomization as a Deployment Tool

In Chapter 3, domain randomization was introduced as a way to improve sim-to-real generalization. Topic 1 reframes it as a deployment practice:

• Randomize friction coefficients, lighting, sensor noise, and obstacle placement in simulation.
• Run regression suites across this randomized space before approving a new firmware or software release.

You will design:

• Simple parameter sweeps that validate whether your balance controller, navigation stack, and manipulation pipelines remain stable across a realistic range of conditions.

Sim-to-Real Transfer and System Identification

You will:

• Compare logs from simulation and field tests for the same trajectories or tasks.
• Use discrepancies (e.g., different stopping distances, overshoot, or slip) to refine your simulation models.
• Learn the basics of system identification for motors and joint dynamics to keep the digital twin in sync with aging hardware.

Calibration of Sensors and Articulated Joints

Finally, Topic 1 establishes calibration as a routine, repeatable process, not a one-off:

• Camera and depth sensor calibration (intrinsics and extrinsics).
• IMU alignment and bias estimation.
• Joint zeroing procedures and kinematic validation tests.

You will prepare:

• Calibration checklists that must be completed before a new site deployment or after major maintenance.