Twipe Robot Eco System
The Twipe robot end to end architecture model is based on a five layer IoT technology stack.
In this article we will review the high level details of this architecture.
Layer 1: Device Hardware
The robot has a two wheeled inverted pendulum design. It uses a custom PCB to integrate a series of off the shelf development boards to dynamically balance the inherently unstable robot.
Main processor
The processing logic for the robot is provided by an Adafruit Huzzah32 development board which features an Espressif ESP32 System On A Chip (SOC). This processor is responsible for handling the logic to balance the robot as well as to handle all the real-time communication with an operator as required. We have an article that described the ESP32 SOC in greater detail if you are keen to know more about it.
Motion Sensors
The robot’s motion tracking is provided by an Invensense MPU6050 inertial measurement unit development board which combine a 3-axis microelectromechanical gyroscope and a 3-axis accelerometer on the same silicon die, together with an onboard Digital Motion Processor (DMP), which processes complex 6-axis MotionFusion algorithms.
Power Management
Twipe is powered by direct current provided by a Mastercraft 20V max lithium-ion battery. This voltage is stepped down to a 12 volt bus, a 5 volt bus and a 3.3 volt bus using a Drok LM2596 multiple output power supply.
Motor
Twipe moves on two Makeblock 42BYG bipolar stepper motors driven by signals from the ESP32 SOC via a pair of Polulu MD20B development boards featuring the Texas Instrument DRV8825 bipolar stepper motor driver.
Physical Properties
The Twipe robot’s chassis measures 8.5” tall by 5.25” wide by 3.75” deep. When the wheels are attached to TWIPe the robot’s height measures 8.625” tall. Twipe’s body is milled from Acetron GP and weighs about 2312 grams (including the battery) and required a calculated detent torque of 0.073 Kg-m in order to maintain its balance in a fixed position. See Doug’s article for more details regarding calculating torque. The robot will need to make aaa DMP calculations per second, update the stepper motors 10 times per second,and send telemetry data to an operator at a rate of ccc bps over a wifi link.
Layer 2: Device Software
The Twipe robot must read accelerometer and gyroscope data from an inertial measurement unit to determine the robot’s orientation and then use a Proportional Integral Derivative (PID) algorithm to maintain an upright posture. Realtime telemetry must be sent out and realtime operator commands must be recieved over a Wifi connection.
Operating System
FreeRTOS is used to run a WiFi stack as well as the primitive xTaskCreatePinnedToCore() which is used to manage mutli-threaded process execution.
Application Software
Embedded systems are programed in C++ using Arduino core libraries.
Layer 3: Communications
This section outlines the communication implemetation strategy used by the Twipe robot.
Infrastructure
Twipe features a 2.4GHz WiFi radio, a WiFi access point, an MQTT broker and a client device capable of running a web browser.
Identification
The unqiue media access control address of the radio prefixed by ‘twipe’ is used to uniquely identify each Twipe robot to both the required WiFi Access Point and the MQTT broker. This makes it possible for multipe robots to co-exist in the same network.
Data Protocols
Communication between the robot and the MQTT broker is comprised of MQTT messaging on port 1883 encapsulated in IPv4 packets over a 2.4GHz 802.11n WiFi connection at 150Mbps. Communication between the MQTT broker and the web browser client is comprised of MQTT messages on port 1883 tunneled through a websocket connection on port 9001 encapsulated in IPv4 packets over whatever speed the local Access Point supports.
Device Management
Monitoring and dynamic configuration of the robot is done via MQTT messages between the web browser client and the robot. While not yet implemented the plan is to include Over-The-Air (OTA) for coe updates. Today code updates are done via a local USB micro conection.
Layer 4: Data Analytics
The robot outputs two types of data. One is health data in the form of error counts for various events of interest. The second is telemetry data which provides real time sensor data regarding the balancing activities of the robot.
Layer 5: Applications
The Twipe robot runs a monolithic firmware image written in C++ and located in a Github respository. The operator console runs in a web browser and is written in HTML, Javascript and cascading style sheets.