Decisions to avoid collisions

The National Research Council of Canada supports detect-and-avoid research to improve drone safety

July 17, 2018— Ottawa, Ontario

Technology advances in unmanned aerial vehicles— also known as drones— are pushing beyond recreational use and infrastructure monitoring to transportation, parcel delivery and other applications. The development of drone prototypes small enough to fit in the palm of your hand or large enough to operate as an air taxi is creating an urgent need for technology solutions that will allow their integration into our already crowded airspace. The National Research Council of Canada (NRC) is studying, developing, and testing detect-and-avoid technologies to make advances on preventing collisions between drones and other aircraft.

Stand-in drone concept brings new testing approach

For routine drone operations to occur beyond a pilot's visual line of sight, several technological solutions need to be developed and tested. The drone must be able to communicate with others using and monitoring the airspace, detect other aircraft, and use intelligent decision-making technology to avoid a crash. These detect-and-avoid functions must also work on their own when the control link to the remote drone pilot is not available.

NRC researchers are taking a new approach to testing detect-and-avoid systems by using the NRC's Bell 205 Airborne Simulator Aircraft. Equipped with computer-controlled automation that operates the helicopter autonomously, the Bell 205 acts as a stand-in for a drone with a safety pilot on board to assume control if needed. The surrogate drone concept has several advantages for flight testing:

  • It can carry much heavier cargo than small drones, so the equipment does not need to be made smaller before it can be tested in the air;
  • The presence of a safety pilot allows for the testing of systems and equipment that are not fully developed, without risking the loss of a drone. This approach provides access to airspace without the additional approvals needed to test a drone, and allows for testing under real airspace operating constraints due to nearby traffic at the Ottawa International Airport. As a result, the test data reflects a realistic operating environment.

Flight testing innovative collision avoidance technology

Automatic collision avoidance technologies provide the artificial intelligence a drone needs to recognize that another aircraft is a potential crash risk and change course to prevent getting too close. This technology must identify a threat and make decisions in a verifiable way in a given period of time. NRC researchers have developed and flight-tested a collision avoidance algorithm that uses a dynamic model to determine the latest possible moment to alter a drone's path and still guarantee it will avoid hitting another aircraft.

Based on continuously evaluated sensor data, the algorithm works through a series of decisions about possible collision threats, and how and when to change direction. If a threat is detected, a warning is sent to the drone pilot with a preview of the drone's automatic response in case the pilot is out of communication or unable to take control, leaving the drone to respond independently. The algorithm factors in uncertainty regarding the threat situation and uses a decision tree to determine the most appropriate manoeuvre. In doing so, it must respect the 'rules of the air' and consider the manoeuvering limits of the drone. It must also delay manoeuvering as long as possible, as the sensor data is more reliable with proximity to incoming aircraft, and because it allows the human pilot time to resolve the situation using knowledge that may not be accessible to the automation— for example, local airspace procedures or routes, information about the other aircraft's flight path, and radio communications with the other aircraft.

In October 2017, after developing the means to conduct near-miss flight trajectories for the study, NRC researchers put the algorithm in the air to test it using the NRC's Bell 205 helicopter as a stand-in drone, and the Harvard Mark IV as an 'intruder.' While the two aircraft flew from beyond visual detection range toward a potential crash point, the algorithm detected the intruder, identified the collision threat, and made navigational decisions that delayed manoeuvering until the last possible moment. The algorithm accomplished all this while still guaranteeing a pre-set 'miss distance,' which allowed the researchers to test the fully autonomous collision avoidance mode. All six flight tests were successful, and intruder pilots noted the manoeuvres made were appropriate.

In summer 2018, the researchers will return to the air to test the technology again, this time relaxing some rules in the decision tree to evaluate a scenario in which the incoming aircraft is detected later. They will gather physiological data and subjective feedback from pilots to see how they react to a drone behaving in a way that may not be expected, and to shed light on the implications of automation that has the same safety goals but may obey different rules. The system prototype, complete with algorithms, is being readied for any interested manufacturers that may want to license its supporting technologies to incorporate into a product.

Refining and proving ground-based detect-and-avoid technology

NRC researchers are working with Seamatica Aerospace and Drone Delivery Canada (DDC) to test a mobile, ground-based, detect-and-avoid radar system that scans the sky for aircraft and warns a drone pilot when it detects one. Having the sensor on the ground can be better for certain drones and missions, such as operations within 10 kilometres of the sensor, or when using drones that are less able to carry and power an on-board detect-and-avoid system.

With DDC operating their drone and the NRC deliberately crossing the drone's path with its Twin Otter research aircraft, researchers are helping Seamatica fine tune their radar and conflict detection software. Testing with the drone operating within a pilot's visual line of sight is complete; testing for operation beyond visual line-of-sight will take place at a test range near Ottawa. Once proven there, the system will be deployed in an off-range test area as one of Transport Canada's Beyond Visual Line-of-Sight Pilot Projects, where it will support the demonstration of safe cargo deliveries to remote Canadian communities.

Supporting industry and regulators

The NRC plays a key role in detect-and-avoid technology because of its considerable expertise in flight research and testing, sensors, and system integration and its role as a trusted third party to manufacturers and regulatory agencies. Since 2007, the NRC has participated in Canada's efforts to develop drone rules by analysing proposed operating standards and regulations based on research into the capabilities and deficiencies of drone technologies. The NRC also helped advance drone research by making its flight test video data publicly available, including to universities and researchers in collision avoidance systems.

For more information about the NRC, its aerospace programs, capabilities and facilities, visit our Aerospace research and development expertise website.

Additional Links

Civilian Unmanned Aircraft Systems program

Media enquiries

Media Relations, National Research Council of Canada
1-855-282-1637 (toll-free in Canada only)
1-613-991-1431 (elsewhere in North America)
001-613-991-1431 (international)
media@nrc-cnrc.gc.ca
Follow us on Twitter: @NRC_CNRC

Stay connected

Subscribe

Date modified: