The requirements below are derived from an objective tree and use cases that take into consideration the project goal and sponsor expectations.
Functional Requirements
The System shall:
- Autonomously move
- Detect obstacles
- Classify obstacles
- Predict the path of moving objects
- Autonomously avoid obstacles based on classification
- Operate in real-time
Mandatory Performance Requirements
The mandatory performance requirements are divided between the virtual and real-life systems. The virtual system’s requirements are derived from the characteristics of the sponsor’s robots and the end goal of the project.
The virtual system will:
M.P.1 Autonomously move with max speed of 1.8 m/s
M.P.2 Autonomously move with max rotation speed of 60 deg/s
M.P.3 Autonomously move with min cruise speed of 0.5 m/s
M.P.4 Increase robot productivity by > 5% using avoidance strategies based on object classification, compared to the nominal avoidance(Local A*)
M.P.5 Operate in real-time with planning time within 0.1 s
Requirements M.P.1 through M.P.3 are based on the hardware specifications of the sponsor’s physical robots. M.P.4 will be used to validate the project’s high-level goal, which is to increase productivity. M.P.5 is included to ensure the system would be feasible for real-time operation.
The real-life system’s requirements are derived by scaling the projected metrics in a full-scale environment down to the scale of the project’s real-life system. The performance metrics are also based on additional trade studies and discussions with the sponsor.
The real-life system will:
M.P.6 Classify obstacles with precision of 70%
M.P.7 Classify obstacles with recall of 80%
M.P.8 Detect center positions of obstacles of interest within 0.1 m accuracy
M.P.9 Detect model forklifts within a range of 2 m
M.P.10 Detect model pedestrians within a range of 1 m
M.P.11 Output results of positioning and classification within 100 ms per frame
These requirements assume that the project’s real-life system is scaled down from the full-scale environment by a factor of 10. This scale was chosen based on testing locations available to us and discussion with the sponsor. Requirements M.P.6, M.P.7 and M.P.8 will be used to validate obstacle classification and localization performance, respectively. M.P.9, M.P.10 and M.P.11 are based on the expected required detection range and computation time of the robot needed for successful collision avoidance.
Mandatory Non-Functional Requirements
The non-functional requirements do not feature numerical metrics; instead, they qualitatively capture some of the system characteristics requested by the sponsor.
Both virtual and real-life systems shall:
M.N.1 Use mobility subsystems that are non-holonomic
The virtual system shall:
M.N.2 Have modular avoidance subsystems
M.N.3 Be able to operate in a non-empty environment
M.N.1 matches the kinematic restriction on the sponsor’s current robots. M.N.2 is included to allow comparison against the nominal avoidance method (local A*) to show productivity gains. M.N.3 simply ensures the system’s test environment is not trivial or completely empty.