e22-3yp-Syncrobots

Project Title

---

Team

• e22044, Nithika Bandara, email
• e22135, Thimira Dilshan, email
• e22353, Gayasha Sandeepa, email
• e22421, Tharindu Weerasinghe, email

Table of Contents

1. Introduction
2. Solution Architecture
3. Hardware & Software Designs
4. Testing
5. Detailed budget
6. Conclusion
7. Links

Introduction

Coordinated multi-robot formations are needed in warehouse logistics, precision agriculture, environmental monitoring, and search-and-rescue. Existing demonstrations usually depend on motion-capture systems such as Vicon or OptiTrack, or on GPS. These are expensive, limited to controlled spaces, or unavailable indoors, so swarm robotics research stays out of reach for student groups and small organizations.

This project builds a low-cost swarm that measures its own relative positions and holds a commanded formation in an ordinary indoor space, with no external infrastructure. Each robot carries an Ultra-Wideband (UWB) ranging module that gives centimeter-level distance measurements without cameras, markers, or GPS.

A simple leader-following chain suffers from string instability, where oscillations grow along the chain. To reduce this, each follower tracks its distance to both the global leader and its immediate predecessor. The result is a decentralized, infrastructure-free formation controller built from inexpensive hardware, so more groups can experiment with swarm robotics.

Solution Architecture

Topology. One leader (R0) and several followers (R1, R2, …). Each follower Ri holds two distance constraints: one to R0 and one to R(i-1).

On-board data flow (every 20 Hz control cycle):

1. UWB SDS-TWR ranging produces raw distances to the leader and the predecessor.
2. A 1D Extended Kalman Filter suppresses the 5 to 15 cm noise and multipath spikes.
3. A potential-field controller combines the filtered distances with the heading estimate from the IMU and wheel encoders.
4. A unicycle mapping converts the controller output into left and right wheel commands, which drive the motors by PWM.

Communication.

• Ranging: Symmetric Double-Sided Two-Way Ranging (SDS-TWR), which cancels clock drift without synchronized clocks.
• Scheduling: a 50 ms TDMA frame gives each robot a 10 ms slot, so the whole swarm refreshes at 20 Hz.
• Telemetry: the leader sends periodic summaries over ESP-NOW to a base-station laptop, which logs distances, control errors, and battery voltage to CSV.

Decentralized control. Every follower computes its own steering, so the swarm keeps running if the base station goes offline. Only the desired formation distances are set globally.

Security. ESP-NOW is unauthenticated by default, so each robot keeps a whitelist of paired MAC addresses and discards packets outside its TDMA slot. No personal data is collected, only robot telemetry.

Hardware and Software Designs

Hardware

All nodes use the same bill of materials, so any unit can act as the leader or a follower.

Function	Component
Microcontroller	ESP32 NodeMCU (30-pin)
UWB ranging	Decawave/Qorvo DWM1000 / DW3000 module (SPI)
Heading	MPU6050 6-DoF IMU and N20 magnetic encoders
Drive	2 x N20 6 V gear motors, MX1508 dual motor driver
Chassis	2-wheel round double-deck smart car chassis, 65 mm wheels
Power	3S 18650 pack, 3S BMS, LM2596 buck (6 V motor rail), Mini-360 buck (5 V logic rail)
Level shifting	4-channel 3.3 V to 5 V logic level converter

Power architecture

The battery pack feeds the BMS, then a rocker switch, then two buck converters. The LM2596 supplies the 6 V motor rail to the MX1508 driver. The Mini-360 supplies 5 V to the ESP32, whose onboard regulator provides 3.3 V for the UWB module and IMU, since neither is 5 V tolerant. All grounds are common.

Software

• Firmware (ESP32): UWB driver and SDS-TWR exchange, TDMA scheduler, Kalman filter, potential-field controller, unicycle mapping, motor PWM, and ESP-NOW with MAC whitelisting.
• Antenna delay calibration: each UWB module is calibrated against a known distance before use.
• Base station: a script that receives telemetry, writes CSV logs, and plots distance-tracking error and formation shape.

Testing

Test	What it checks
UWB ranging accuracy	Measured vs. true distance at several known distances, before and after antenna delay calibration
Range noise and filtering	Raw vs. Kalman-filtered distance noise on a stationary pair
TDMA timing	Whether every robot gets its ranging slot at 20 Hz with no collisions
Heading estimation	Yaw drift from gyro and encoder dead reckoning over a run
Single-follower tracking	Leader moving at up to 0.3 m/s, tracking error of one follower
Full formation	Leader plus two followers holding the commanded formation
Battery life	Run time of continuous formation driving per charge
Link loss	Whether followers keep operating with the base station switched off

Detailed budget

All items and costs

Item	Quantity	Unit Cost	Total
Sample item	5	10 LKR	50 LKR

Conclusion

What was achieved, future developments, commercialization plans

Links

• Project Repository
• Project Page
• Department of Computer Engineering
• University of Peradeniya