Peppa Punch combat robot

The brief

Design, build and fight a featherweight robot in the module’s Robot Wars competition, under the Fighting Robots Association build rules and the university’s own constraints: no more than 13.6 kg, small enough for a university locker, an approved radio, a removable link or kill switch, at least 3 minutes of continuous running under combat conditions, and a pass at safety inspection.

Every team got the same kit: two 18 V drill motors with gearboxes, two 18 V 1.5 Ah Li-ion drill batteries and a radio set. For motor control we could choose a servo-operated gear selector, a pre-programmed relay board, or an H-bridge built from scratch.

The team and my role

There were seven of us, and I was the group leader.

Leading the team

I allocated roles across CAD, machining, welding and electrical sub-teams, and kept communication going within the team and back and forth with our lecturers and the lab technicians.

Project management

I planned the build on a Gantt chart in ProjectLibre, from defining the concept on 10 March to the competition on 10 May 2026, and tracked it in weekly meetings.

CAD and welding

I led the CAD and welding work: I modelled components in SolidWorks, led the wheel design and welded the frame. I also built the shell and spike assembly (28 March to 6 April) and reinforced the chassis and armour (21 to 25 April).

Concept

Each of us proposed a concept, and we scored the seven against eight weighted criteria. “Punch Turtle” won: a high, domed body based on a tortoise shell, for strength and the chance of self-righting, with a defensive strategy of out-sizing opponents and resisting flippers.

(opens full size)

The first design carried a rotating spike on the front. We dropped it because it vibrated, risked serious mechanical failure and was hard to mount on the motor, and the weapon motor’s gearbox was broken during the build. The final weapon was a set of fixed steel spikes.

(opens full size)

Design

Frame

A welded skeleton of 16 mm mild-steel square hollow tube, 400 × 380 × 215 mm, with angled members forming the dome and extra bars either side of the front and back.

Shell

Sheet-steel plates fixed to the base and the frame bars, with recycled aluminium cans as lightweight shrouds, a hinged flap for access to the electrics, and an MDF base plate.

Spikes

Four large spikes (70 mm long, 25 mm in diameter) at the front and nine small ones (40 mm long, 12 mm in diameter) along the sides, lathe-turned to a 20° taper, centre-drilled 20 mm deep and tapped M6.

Drive

Two 3D-printed PLA+ drive wheels and two scooter wheels as front castors, with skid steering. I designed each wheel as a keyed hub and tyre that bolts to the motor’s 50 mm mount with four M5 bolts on a 35 mm square. We doubled the wheel diameter from 60 mm to 120 mm when we found that half of each wheel sat inside the shell and we hadn’t allowed for manufacturing tolerances.

Electrics

The two drill motors (37 N·m maximum torque) ran through the relay board, giving reverse, half speed and full speed forward, from the two battery packs in parallel.

(opens full size)

Build

Frame

We cut the bars on a chop saw and the angles by hacksaw, marking them with a protractor, and tack-welded the frame first. A miscommunication between machining and welding meant several bars were cut at the wrong angles. Because the joints were only tacked, we could grind them off, bridge one gap with a welded rod, re-cut the bars and get the intended dome on the second attempt.

Shell

The sheets had to be bent and sawn to fit the frame’s angles, and the access-flap hinge was reworked until the flap closed.

Spikes

I turned the spikes on the lathe and drilled their mounting holes. Because the sheet was already welded on, the nine side-spike holes were drilled with the whole robot clamped on a pillar drill, and some had to be re-drilled. The 12 mm bar jammed the machine tap, so I started the threads on the lathe and finished them by hand.

Wheels

The university print service printed them after a long backlog (about 8 hours of printing). The 1.5 × 2 mm tread grooves didn’t form on the FDM printer, so we sanded them off and wrapped the wheels in rubber. The first screws tore the rubber, and we refixed it with larger screws.

(opens full size)

Result

We fought three rounds at the competition on 10 May 2026. In the first, a three-way match, the robot got stuck on the arena’s wooden barrier, took damage that disconnected an internal wire, and one drive motor stopped, so we lost. After the repair it ran the final melee with no failures, but with no active weapon and limited pushing force it didn’t win a round. The frame and spikes stayed sound throughout.

Looking back

The welded frame did its job, and tacking before full welding let us recover from the mis-cut bars without starting again. The weaknesses were mostly in the analysis and planning around it.

• The concept matrix didn’t value offence. Punch Turtle scored at least as well as every rival on every criterion, so it would have won under any weighting; the weights decided nothing. The matrix also had no criterion for attacking power, which is exactly what we lacked once the weapon was dropped, and it gave full marks for functionality while the weapon was still undecided.
• Our calculations had errors. The speed estimate used the drills’ 600 rpm no-load speed together with their 37 N·m maximum torque, which a motor can’t deliver at the same time, and it still used the 60 mm wheel after we moved to 120 mm. Pushing was limited by traction, not torque: with about 78 N on the driven wheels and a friction coefficient of 0.8, the wheels slip at about 63 N of push, which uses only a few N·m of the 74 N·m the motors could give. Four-wheel drive, or more weight over the driven wheels, would have done more than any motor change.
• By our own figures the battery couldn’t last the day. Three 3-minute matches at about 27.6 A need about 4.1 Ah, more than the 3.0 Ah from the two packs.
• Wheel radius isn’t ground clearance. Clearance is the height of the chassis underside above the floor, and half of each wheel sat inside the shell, so the real clearance was less than the 60 mm radius. Next time I’d measure it.
• We didn’t test what failed. There was no mass budget and no stress or impact analysis, and the weapon, stability, structural and impact tests on our plan were all suspended. Our bench test reached only 3 A, far below the current in a match. The loose wire that lost the first round is what an impact test, and wiring with strain relief and proper connectors, would have caught.
• Design for manufacture came too late. The weldment drawings had no cut list or tolerances for bars cut by hand at compound angles, and the spike holes were drilled through the finished, skinned robot.
• The schedule. Our final design was a week late, which cost us a week of manufacturing time. I used managed contingency over the Easter break to recover it, and we were ready for the competition, but the build stayed tight. Next time I’d plan the break in from the start, with smaller sub-tasks for each person.

What I’d do differently

• Include offence in the concept scoring and check how sensitive the result is to the weights.
• Keep a mass budget from the start and weigh every part.
• Give the welders a cut list with every member’s length and angles, use a jig for the angled cuts, and dry-fit before welding.
• Drill the shell sheets from a template before welding.
• Use bought rubber wheels or four-wheel drive, and put the battery weight over the driven axle.
• Fuse the battery, fit a rated removable link and a tested radio failsafe, and use connectorised, strain-relieved wiring.
• Measure speed, current, run time and push force, and run a mock arena test before the competition.