Objectives

Isolated autonomous machines must rely on individual machine sensors with limited range and isolated situational awareness. The result is that these isolated machines must act conservatively and myopically. In the case of automated vehicles, this means cautious, low-speed maneuvering. At a system level this implies a significant reduction in overall capacity and throughput relative to current conditions with human drivers. Passive or isolated autonomy does not scale to the system level for many use cases, including drone swarms and/or automated vehicles.

The objectives of the Noblis Pieces of Eight (Po8) project is to enable nearby connected autonomous machines collectively plan motion paths and other actions that avoid collision or other conflicts. Machines that receive priority compensate machines that give way to ensure equity over time. The complexity of these paths among unfamiliar machines is dependent on an earned trust report (secured via blockchain technology) for each individual machine based on reliability in prior interactions.

Background

Consider an ecosystem of interacting heterogeneous autonomous connected machines, including machines which are both mobile, autonomous, and connected (e.g., autonomous passenger vehicles, autonomous drones, autonomous robots traversing hostile terrain). These machines, all owned by different parties (private individuals, delivery companies, or different law enforcement agencies) may find themselves near one another yet seek to carry out their different intended tasks without conflicting with or colliding with each other. The machines are heterogeneous (that is, different from each other) in terms of design, capability, speed, and independent (passive) sensing ability. Coordinated action among these autonomous machines may be critical – but may have little prior experience coordinating motion and actions.

Limited notions of coordinated autonomy exist, but they are either localized human analogs (e.g,. turn indicators) or depend on 100% trust among machines. Without uniform and systematic method to build up (or reduce) earned trust, there can be no systematic misbehavior detection and adaption. Heterogeneity of isolated autonomous machines at scale is problematic for isolated or top-down control – particularly in systems predicated on binary trust constructs.

Approach

The Noblis Pieces of Eight (Po8) system enables nearby connected machines to share situational awareness regarding obstacles and threats projected over time, and collectively plan motion paths and other actions that avoid collision or other conflicts. This coordination permits high-speed motion and coordinated actions impossible without advance knowledge of all coordinated localized actions.

Further, the Po8 System enables a collective post-hoc accountability process to assess the reliability of each individual machine to act faithfully in accordance with collectively optimized motion paths and actions. An individual machine establishes a track record within the Po8 System (secured using a blockchain) so that the reliability of the machine to faithfully follow collectively assigned motion/action paths in the past may be factored into the representation of the collective uncertainty map and the current set of optimized motion/action paths.

The Po8 System has five fundamental elements:

  1. Collective situational awareness, realized in a collectively normalized local obstacle and threat map (uncertainty contour) normalizing diverse sensory input from heterogeneous machines and diverse sensor technologies into a single shared local dynamic map that explicitly considers the past reliability (earned trust) of individual machines.

  2. Optimization of collective motion/action traversing uncertainty contours, reconciling potential conflicting motion paths or actions among all machines and cognizant of the reliability (earned trust) of individual machines.

  3. Value distribution by contribution, a method for equitable remuneration from machines with advantageous path/action allocations to local machines for contributing to uncertainty reduction and lo-cal/non-local machines contributing to the administration of Po8 System functions.

  4. Post hoc individual accountability assessment, in which all machines assess the ability of nearby machines to act and move in accordance with collectively determined motion/action paths;

  5. Earned trust accounting and reporting, for each individual machine, in an open consortium side chain distributed ledger – to ensure transparency, accuracy and security of the Po8 system.

Repository Access

The code has not been released publicly. Private access has been provided to the MOBI challenge judges.

Built With

Share this project:
×

Updates