Stay up to date with the latest developments from the MCAV team by reading about our current projects below.
The MiniDrone is a small-scale version of the autonomous vehicle, designed for educational purposes and testing of autonomous features. The goal of the project is to provide an affordable and approachable platform for developing and learning about autonomous vehicles.
The hardware section aids the accomplishment of this goal through the development of a robust chassis to protect the computation hardware and power systems, whilst maintaining ease of access to the aforementioned components. A later stage version of the chassis is depicted, with construction and testing scheduled for completion before the teams end of year demonstration.
More information on the project can be found at: https://sites.google.com/student.monash.edu/minidrone/home
Armoured Dynamic Obstacle
The Armoured Dynamic Obstacle (ADO) is a user-controlled device whose purpose is to facilitate testing of the team's autonomous vehicles. This is achieved by being able to mount cyclist and pedestrian dummies to the chassis in order to simulate real-world scenarios. The design goal for the chassis is to be able to withstand the impact of a collision with the team’s electric vehicle, whilst protecting the internal electrical components. The ADO allows the team to evaluate and test its object detection algorithms and verify successful emergency stopping, as well as other evasive manoeuvres. The team is scheduled to complete the design by the year’s end, with construction occurring early 2022.
The team is currently in the process of redesigning the actuation systems retrofitted to the team’s Subaru Forester. The team have critically analysed the existing braking and steering actuation systems to iterate and further improve upon the design. Key to both of these actuation systems is safety, with manual overrides and the capability for the driver to assume control over the vehicle at any point in time. The vehicle is intended to operate in a connected environment. As such, the team's goal for 2021 is to have an on-board unit (OBU) inside the vehicle receive a command to perform an emergency stop, and be able to safely and successfully respond to this command.
This year the Interactions Section have been working on projects to further MCAV's investigation into the interaction between autonomous systems and elements in its surroundings. In particular, our section's focus this year was the interaction between autonomous systems and general traffic (comprised of private cars) and the interaction between autonomous systems and human operators/passengers.To study the interaction between autonomous systems and general traffic, we have used SUMO to simulate traffic and intersection control at a 4-way intersection. Using Sidra Analysis, we deduce and design an optimal sequence for intersection control of autonomous systems. The limits of this sequence is tested in the SUMO environment - where we also test the sequence's performance under high congestion events such as roadworks or incidents occurring at the intersection.The traffic simulation developed in SUMO is used to generate traffic in our SUMO/CARLA 3D co-simulation environment. This environment provides an immersive tool to test our autonomous system and its serviceability for human operators and passengers. We are currently developing a project with external members to integrate this co-simulation with VRAV technology — which allows us to test our systems and their impact on human operators and passengers in a more immersive manner.
Adaptive Cruise Control & Emergency Braking
The Lane Keeping system controls the steering of the car, but does not react to other cars. This is where our new Adaptive Cruise Control system will come in. ACC will be responsible for regulating the speed of the car. When active, it will maintain speed at the speed limit if possible, but slow down for other vehicles and then speed back up again.
The LiDAR sensor on the StreetDrone will be responsible for detecting other vehicles and obstacles. These detections will be taken into account by the controller which will determine a smooth but safe amount of acceleration to apply. The ACC system will also be responsible for applying brakes in the case of an imminent collision.
Lane Keeping Assist
A lane-keeping system controls the steering of a car to ensure that it stays between the lane lines without requiring constant input from the driver. This reduces the cognitive load on the driver, making driving a more pleasant experience, and represents one step towards fuller autonomy.
In 2020, we created a neural network that takes a camera view from the car and separates the image into segments showing the lane that the car is in. We added a controller onto this that steers the car to keep it within the current lane, resulting in our first lane-keeping system that we trialled in simulation.
We are building on this system in 2021, where we hope to improve performance and robustness and plan on testing the system in the real world on the StreetDrone.
The MiniDrone is a small-scale autonomous vehicle focused on education and testing. The goal of the project is to provide an affordable and approachable platform for developing and learning about autonomous vehicles.
Development began in Semester 1 2021, where we took a 1/16th scale RC car and added a GPS, an IMU, a camera, a 2D laser scanner and a Jetson Nano computer. This served as the initial prototype for software development. In parallel, the team worked on designing a slightly larger, more robust and realistic vehicle from scratch which is due to finish manufacture early in Semester 2.
The MiniDrone has already proved effective for demonstrations and tutorials and we hope that it can be an effective option for testing self-driving algorithms in a low-stakes environment before tackling implementation on the StreetDrone.
The Connectivity team are currently working on accuracy for the DSRC geolocator, researching how to implement our own cellular network, that is using a Jetson hat. The ICA algorithm has moved to a V2V solution. The algorithm has been improved with more research on different scenarios and previously used algorithms. We are also working with the interactions team to test out traffic flow simulations and use those to understand traffic response when the algorithm is used and smooth out any design flaws. Also the connectivity architecture has been changed with the Jetson Nano overlooking the DSRC and the cellular infrastructures.