Abstract
Over the past decade, self-driving capability for all variants of on-street vehicles have promised safer and more efficient transportation. This remains “work in progress” with large unfilled gaps in addressing user-acceptance, safety, ethics, regulation, technology and the business model. Our goal is to develop the Open-source Autonomous Vehicle (AV) software for Open-standard Electric Vehicle (EV) platforms, ie. AV4EV paradigm, to help realize safe, reliable, and efficient autonomy for off-street use cases. In particular, we focus on developing the AV4EV Autonomy Essentials Kit (AV4EV-Kit) for known controlled application domains: logistics (in-warehouse mobile robots), material handling (autonomous forklifts) and airside cargo (autonomous ground support equipment). The AV4EV business model addresses these many smaller domains through simplification and modularity. The EV ‘skateboard’ chassis is orders of magnitude simpler than on-street vehicles (~20 moving parts compared to nearly 2,000 in contemporary vehicle architectures) - supporting standardization of interfaces for autonomous driving. Modularity allows AV4EV to address autonomous vehicle market sizes of 50K-250K vehicles/year for each use case by enabling component re-use and efficient customizability to meet specific segment needs. If successful, the AV4EV Kit will create a new business category for Autonomy-as-a-Service with plug-n-play hardware and software for rapid prototyping and deployment. Autonomous machines have a serviceable market of $2.9B with a 15.5% growth rate.
The AV4EV Autonomy Essentials Kit enables logistics customers to kickstart their journey of autonomous machines for safe and efficient movement of people and goods, even if their companies have little prior autonomous system development experience. Using the AV4EV-Kit, customers can rapidly prototype EV platforms into autonomous machines in 10 days for brownfield deployments.
The AV4EV Autonomy Essentials Kit is dedicated to lowering the entry barrier of autonomous driving development and deployment. AV4EV-Kit consists of (1) a plug-in-play hardware platform with sensors and compute, (2) an autonomy software stack to achieve essential autonomous driving functions of perception, sensor fusion, mapping, localization, path planning, obstacle avoidance, traffic light recognition and safe control; and (3) a new Software Defined Vehicle approach for autonomous machine software development and testing in the cloud to lower cost of mixed-criticality software and over-the-air upgrades to enhance safety across the vehicle lifecycle and customize for different deployment scenarios. The AV4EV-Kit conforms to the open-source Autoware autonomous vehicle software standard to interface with the EV’s drive-by-wire system for users to easily integrate navigation functions with vehicle control. The AV4EV-Kit incorporates energy-efficient machine learning-based perception, planning and control algorithms developed by the PI’s and Co-PI’s labs and will be tested by commercialization partners on a variety of EV platforms.
Description
Timeline
Strategic Description / RD&T
Our proposed work is to develop a modular open-source AV4EV Autonomy Essentials Kit with the necessary hardware options and autonomous driving software capabilities for emerging standardized electric vehicle skateboard chassis in off-street use cases in transportation and logistics. Unlike high-volume mass produced on-street vehicles which require an optimized design and manufacturing process to realize economies of scale, for off-street, our target autonomous machines markets are smaller - with volumes of 50K-200K vehicles/year for each use case. So modularity and reusability provide efficient customization for meeting customers’ needs quickly. If successful, the AV4EV Kit will create a new business category for customizable Autonomy-as-a-Service with plug-n-play hardware and software for rapid prototyping and deployment. AV4EV sees an opportunity to catalyze this alternative path to adoption of next-generation autonomous machine technologies, based on 3 mutually-reinforcing strategic pillars:
1. Focus on narrower, well-defined and understood solution domains in logistics, cargo and material handling rather than general-purpose vehicles.
2. Combine AV and EV in a common solution architecture, reducing overall vehicle part count and complexity, while unlocking benefits of engineering the technologies together.
3. Start with a modular, Software-Defined Vehicle architecture, enabling common hardware and software components to be used across many different solution domains, while facilitating design integration, configuration and validation of domain-specific combinations of components in the cloud.
Deployment Plan
Milestone 1 (Year 1 - 2nd/3rd Qtr): (A) AV4EV-Kit requirements: outline requirements for the cargo ODD with Platform-0 AV4EV Go-kart. Develop operational design domain or driving scenario (ODD) and configure sensor stack. (B) Software Define Vehicles: identify first use case for mapping Go-Kart computation into AWS and demonstrate over-the-air updates. (C) Acquire Platform-1 from EveAutonomy and interface it with the sensor stack.
Anticipated Outcome: Demonstrate Cargo driving scenario with AV4EV Platform-0 Go-Kart
Milestone 2 (Year 2 - 1st/2nd Qtr): (A) AV4EV-Kit requirements for indoor and outdoor use case with pedestrian safety requirements. Build and demonstrate the sensor stack with safe control over the near term 2-3 seconds horizon. (B) Software Define Vehicle - demonstrate complete flow of multiple electronic controller unit code development, testing and deployment in AWS. Demonstrate machine learning model on-line update. (C) Acquire Platform-2 from PixMoving (Autoware partner) to demonstrate portability of AV4EV-Kit. Conduct site visit with Autoware Foundation to understand the airport-specific ground equipment transportation ODD requirements. In addition, development partners will be identified from the focus group sessions with the Autoware Center of Excellence (CoE) members. They will meet quarterly with Autoware Foundation to address development issues with the robot. This team will function as the early assessors and will provide real time feedback.
Anticipated Outcome: Update AV4EV stack from prototype version to pre-production version that is designed for manufacturability. The goal is to scale the production for the AV4EV-Kit. Develop Go-to-Market plan with Mack Institute.
Milestone 1 (Year 2 - 3rd Qtr): Complete integration of AV4EV-Kit and Software Defined Vehicle within Autoware with a configuration and deployment tool for rapid prototyping on customer sites.
Anticipated Outcome: Demonstrate fully-operational system over multiple days of operations at Autoware Foundation sites. Accompany them on industry roadshows to capture pilot customers. Engage with Mack Institute to meet with investors and create customer contracts and CoE to complete purchase agreements.
Expected Outcomes/Impacts
1. Demonstrate Cargo ODD with AV4EV Platform-0 Go-Kart
2. Update AV4EV stack from prototype version to pre-production version that is designed for manufacturability. The goal is to scale the production for the AV4EV-Kit. Develop Go-to-Market plan with Mack Institute.
3. Demonstrate fully-operational system over multiple days of operations at Autoware Foundation sites. Accompany them on industry roadshows to capture pilot customers. Engage with Mack Institute to meet with investors and create customer contracts and CoE to complete purchase agreements.
Expected Outputs
AV4EV will create research prototype reference platforms for academic research in developing autonomous vehicle software and systems for open-standards electric vehicle platforms.
The AV4EV Autonomy Essentials Kit is dedicated to lowering the entry barrier of autonomous driving development and deployment. AV4EV-Kit consists of (1) a plug-in-play hardware platform with sensors and compute, (2) an autonomy software stack to achieve essential autonomous driving functions of perception, sensor fusion, mapping, localization, path planning, obstacle avoidance, traffic light recognition and safe control; and (3) a new Software Defined Vehicle approach for autonomous machine software development and testing in the cloud to lower cost of mixed-criticality software and over-the-air upgrades to enhance safety across the vehicle lifecycle and customize for different deployment scenarios. The AV4EV-Kit conforms to the open-source Autoware autonomous vehicle software standard to interface with the EV’s drive-by-wire system for users to easily integrate navigation functions with vehicle control. The AV4EV-Kit incorporates energy-efficient machine learning-based perception, planning and control algorithms developed by the PI’s and Co-PI’s labs and will be tested by commercialization partners on a variety of EV platforms.
TRID
Building a community to further the development of autonomous systems on electric vehicle platforms is a new direction. While private parties and companies have been developing proprietary solutions fixed to a single platform, our goal is to develop, distribute and deploy open-source software for open-source EV designs for prototyping and to facilitate research in development of safe autonomous vehicles.
Individuals Involved
| Email |
Name |
Affiliation |
Role |
Position |
| rahulm@seas.upenn.edu |
Mangharam, Rahul |
University of Pennsylvania |
PI |
Faculty - Tenured |
Budget
Amount of UTC Funds Awarded
$100000.00
Total Project Budget (from all funding sources)
$100000.00
Documents
Match Sources
No match sources!
Partners
| Name |
Type |
| The Autoware Foundation |
Deployment Partner Deployment Partner |
| Wharton School Mack Institute |
Deployment & Equity Partner Deployment & Equity Partner |
