Pennsylvania Autonomous Vehicle Policy Task Force Safety Guideline

The National Highway Traffic Safety Administration estimated that 37,461 traffic fatalities occurred last year. Human error is attributed to roughly 94% of vehicle crashes. Removing this human element is a key reason why driverless vehicles are expected to someday prevent thousands of injuries and deaths from accidents.

Highly automated vehicle (HAV) technology is actively evolving across Pennsylvania. The state’s Autonomous Vehicle Policy Task Force has formally implemented their policies regarding the testing of these vehicles. The group has collaborated with a broad range of experts from the state, federal, and private sectors including AAA, the Federal Highway Administration, Carnegie Mellon University and General Motors.

Key Testing Requirements

• A proposal for HAV testing must be submitted and approved by the Pennsylvania Department of Transportation (PennDOT) and compliance with related state and federal standards for safety must be confirmed
• PennDOT must receive notification of intent to use vehicles in a “driverless” mode (without an occupant)
• The Pennsylvania Turnpike Commission and PennDOT may temporarily limit the operation of HAVs on particular roads
• HAVs must have the capability of recording event data that may be provided to PennDOT if a crash occurs
• HAvs must have current cybersecurity defenses in place
• PennDOT may request a financial summary associated with the HAV testing
• The State Police & PennDOT will present an HAV testing report to state legislators annually

HAV Accident Reporting

Pennsylvania law defines a reportable accident as one that leads to an injury, fatality, or damage that renders a vehicle unsafe to operate. Crash reports do not require that those involving HAVs be specifically identified. Minor accidents are not required to be reported.

Determining Liability

The licensed driver operating an HAV is subject to the laws relating to liability. State agencies have yet to determine the specific provisions regarding liability for when an HAV is operating in a fully automated driving mode. It is likely that traditional tort law and product liability provisions will be used in adopting such guidelines.

HAV Decision Making

What decision would a driverless HAV make if instantaneously a situation emerged that either required striking an oncoming tractor trailer or a guardrail at the edge of a cliff? Vehicle computer-based operating systems will probably react to dangerous situations in ways that are most likely to minimize injuries and damage. An HAV’s perception of reality is based on situational data obtained from sensors, cameras, and infrared devices, which is much different than the senses of humans interacting with their brain.

University of Pennsylvania Testing

Rahul Mangharam is a professor at the University of Pennsylvania’s (Penn) Electrical & Systems Engineering Department. He and six others are using mathematically-based diagnostics to create virtual environments for operating HAV’s. Along with Carnegie Mellon University, they are working on the five-year Mobility21 project, which involves studying and testing the latest in vehicle technology funded by $14 million in federal funds. Scientists from Penn are using software called Computer Aided Design for Safe Autonomous Vehicles that conducts the virtual driving.