all projects

Research

Adaptive Morphing Leg for Search and Rescue

A reconfigurable 5-bar leg that switches between fast traversal and high-force load dragging. Published at ICRA 2026.

ARCLab, UC San Diego · Jun 2024 – Present

Bipedal prototype switching from traversal to load-dragging mode.

Changing link lengths raised measured foot force from 58 N to 106 N on the same motors, and the biped used it to drag a 2.3 kg load.

My role. Mechanical design, prototyping, and experimental validation. Co-first author.

SolidWorks / Ansys FEA / Python / CAN / CubeMars AK60-6

Search and rescue robots have two jobs that pull against each other. They need to cover rough ground fast, then pull a heavy load once they reach someone. A leg tuned for one job is usually poor at the other.

This leg gets both by changing shape. It is a 5-bar linkage whose link lengths change while the robot is running, so the same motors can trade speed for force on demand.

Two modes, one leg

Search mode

Passive links extend and the ground link retracts. The foot can reach farther, clear bigger obstacles, and move faster.

Rescue mode

Passive links retract and the ground link extends. The workspace shrinks into a region of high force, good for short dragging steps.

A gear change alters the torque-speed ratio but not where the foot can reach. Changing link lengths does both.

Results

I measured peak static pushing force at the foot with a crane scale, with motor bus current limited to 1 A and five trials per configuration.

  • Baseline: 58 ± 1 N
  • Retracted passive links: 91 ± 1 N (+57%)
  • Elongated ground link: 106 ± 2 N (+83%)

On the biped, the robot follows a foot path, walks, then reconfigures and drags a load of at least 5 lb (2.3 kg) with the passive links retracted to 18 cm and the ground link extended to 13 cm.

Rescue mode: the legs retract and the robot drags a load.
Simulated workspace force maps and measured foot forces for three configurations
Simulated workspace and horizontal force for each configuration, next to the measured peak forces.

How it’s built

The testbed

A planar testbed with interchangeable link mounts lets me swap in different 5-bar geometries and measure each one. Two CubeMars AK60-6 BLDC motors drive the joints, and I compared measured force against the model’s prediction (τ = Jᵀ·F).

Labeled photo of the planar 5-bar testbed
The testbed, with a crane scale on a rail-mounted platform to read foot force.

Capstan-driven reconfiguration

A worm-gear motor winds a tensioned steel cable around a spool to change a link’s length from 64 to 273 mm.

  • It can’t be backdriven, so the leg holds its shape under load without drawing power.
  • The cable drive is smooth and has no backlash.
Capstan reconfiguration module with steel cable and worm-gear motor
The 1:302 capstan module, with guide bearings and a cable tensioner.

Control and the biped

I wrote the Python position control that runs over CAN and coordinates the joint motors with the gait cycle and the reconfiguration actuators.

The biped has four AK60-6 joint motors, capstan drives for the passive links, and a stepper-driven linear stage for the ground link. It hangs from a boom arm during testing.

Front view of the bipedal prototype on its boom arm
Front view of the biped.
Bipedal prototype dragging a weight
The biped in rescue mode.

What broke

  • The biped was tested on a boom arm, not free-standing.
  • The feet slipped during dynamic load dragging.
  • Terrain interaction and friction aren’t modeled.
  • High-force regions sit near singularities and have to be avoided.

These are the gaps the next stage targets.

Where it came from

The leg started as my senior capstone (MAE 156A and 156B), sponsored by ARCLab. A team of five designed a 5-bar pantograph leg for a hexapod that has to carry a 90 kg person, and built a test bed to compare ways of changing its shape. In simulation, lengthening the ground link gave the biggest and steadiest gain in pushing force.

This paper is the next step: taking that idea from the test bed to a working biped. The full story of the capstone is in Phase 1: Hexapod Leg Senior Design Project.

ICRA 2026

The paper was accepted to ICRA 2026, and I presented the poster there.

What’s next

The single-leg result doesn’t say how the legs should work together with the body. I’ve since optimized the leg’s geometry in simulation, and the next step is the full hexapod. That is covered in Phase 3: Hexapod Rescue Robot.

Presenting the morphing-leg poster at ICRA 2026
At the ICRA 2026 poster session.