The Autoware Foundation (AWF) is excited to announce the release of Autoware.Auto V1.0.0, which was recently used to demonstrate Autoware support for the Autonomous Valet Parking (AVP) Operational Design Domain (ODD). By focusing on the functionality required for AVP, the team was able to validate a wide range of core Autonomous Driving (AD) algorithms and scenarios which will be the basis for support of more complex ODDs in future releases of Autoware.Auto.
The features and capabilities in the release enable Autoware.Auto based Autonomous Vehicles (AVs) to perform:
- Autonomous Parking based on dynamic planning with HD-Maps
- Autonomous Parking using waypoint following without HD-Maps
- Either of the above with and without obstacle detection, avoidance, and stopping
- Integration of the vehicle’s AD system with a web services platform over a wireless network to initiate and manage AVP functionality
Josh Whitley, Senior Software Architect for the Autoware Foundation, coordinated the Autoware.Auto project culminating with the V1.0.0 release, including major software contributions from Arm, Apex.ai, Embotech, Mapless.ai, Open Robotics, Parkopedia, Robotec AI, and Tier IV.
More details about the contents of the Autoware.Auto V1.0.0 release and how to get started with Autoware.Auto can be found at ROS Discourse.
Autoware.Auto ODD Roadmap
The Autoware.Auto Open Source Software (OSS) project will continue to evolve based on the concept of incrementally adding features and functionality targeting well defined ODDs, with the ultimate goal of achieving L4/L5 autonomous driving in any operational environment (e.g. Robo-Taxis operating in dense urban environments). By targeting incrementally more complex ODDs (e.g. higher speeds, traffic density, pedestrians, weather conditions, etc.), the developer community contributing to the advancement of the Autoware.Auto project will be presented with a manageable set of development and verification tasks to address the requirements for the future releases of Autoware.Auto.
Based on this approach, the next ODDs that the Autoware.Auto project will address have been decided to be the Cargo Delivery and Racing ODDs. Cargo Delivery represents a broad range of applications and use cases for autonomous vehicles, typically at lower speeds and in more constrained environments.
Racing has always played an important role in the advancement of automotive technology, and this trend is expected to continue in the area of autonomous driving technology. Safe maneuvering (passing) at high speeds (up to 300kph) in a well constrained environment (race track) will advance the Autoware.Auto AD software stack localization, perception, planning and control functionality.
Beyond the currently selected ODDs, the Autoware Foundation is evaluating the options for future ODDs on the path to achieving L4/L5 AD in complex operational environments based on technical complexity, business opportunities, ecosystem and alliance development.
In addition to the continued advancement of Autoware.Auto, the Autoware Foundation will continue to engage with foundation members and the broader AD ecosystem to integrate their products and solutions into the reference implementations, tools and services which are utilized for development and validation of AD systems based on Autoware. These topics are addressed under the Autoware.IO project.
Stay tuned for updates to the Autoware.org website with more information about the plans for the Autoware.Auto and Autoware.IO projects.
Cargo Delivery, the transport of goods between multiple points or last mile delivery, is a market segment that will grow significantly over the coming years. Adoption and scaling of AD technology in the handling of goods is expected to precede mass adoption in passenger transportation solutions (e.g. Robo-taxis) as the operational environment is much more constrained, and achieving safe operation requires significantly less simulation and real world validation.
Some of the key features and enhancements to Autoware.Auto for support of the Cargo Delivery ODD include:
- Addition of a camera-based object recognition and classification pipeline based on both traditional and neural-network-based image processing
- Addition of a radar pipeline for object detection and multi-radar fusion
- Addition of multi-modal sensor fusion (camera, lidar, radar)
- Support for Global Navigation Satellite System (GNSS) based localization
- Support for Inertial Measurement Unit (IMU) and wheel encoders as odometry sources
- Testing of the e Extended Kalman Filter (EKF) implementation with multiple input sources
- Considerations for various vehicle configurations, such as the towing of one or more trailers and the implications on navigation, planning, and collision detection
Working together with Autoware Foundation members, Autoware.Auto Cargo Delivery capabilities will be demonstrated in use cases such as autonomous baggage handling at airports.
The Autoware Foundation is partnering with the F1TENTH Organization to bring Autoware.Auto to the F1TENTH racing community. The F1TENTH platform is used by many universities as the basis of AD education and for advanced AD research. In addition, it offers students the opportunity to collaboratively compete against other university teams in F1TENTH racing competitions. Students build F1TENTH AVs and learn to develop the AD software which runs on the AV.
In addition, many of these university teams leverage their experience with F1TENTH to compete in full scale Autonomous Racing competitions such as the Indy Autonomous Challenge. However, one of the challenges of moving from 1/10 scale to full scale racing is the re-use of the AD software stack.
By bringing Autoware.Auto to the F1TENTH platform, the F1TENTH Organization and Autoware Foundation will enable the AV developers and racing teams to develop on the F1TENTH platform and leverage significant portions of AD software stack as they move to full scale vehicles.
Some of the key initiatives that the F1TENTH Organization and Autoware Foundation are collaborating on include:
- Porting of the Autoware.Auto software stack onto the 1/10 scale F1TENTH hardware platform
- Enabling of a simulation platform for Autoware.Auto that can be used with both 1/10 and 1/1 scale environments and vehicles
- Development of training collateral and programs for developing with Autoware.Auto on F1TENTH hardware
- Enhancement of F1TENTH hardware, compatible with Autoware.Auto, to enable broader availability and developer community access
- Using F1TENTH AVs as a valuable platform for development and validation of AWF targeted ODDs (e.g. Cargo Delivery)
In addition to enabling the university and developer communities, the F1TENTH Organization and Autoware Foundation plan to host events for simulated and real world racing based on Autoware.Auto. If you are interested in participating in the collaboration between the F1TENTH Organization and the Autoware Foundation, please send an inquiry to firstname.lastname@example.org. Stay tuned for more exciting announcements as Autoware goes Racing!
Latest from the Autoware Foundation
F1Tenth Korea Delegation Visited UPenn CoE
The F1Tenth Global Camp was supported by LINC 3.0, a South Korean government project. LINC 3.0 is
6th Autoware Workshop at IEEE IV 2024, Jeju Island, Korea on Sunday, June 2nd, 2024. Important Da