Ontology-Driven Robotic Specification Synthesis

This paper addresses robotic system engineering for safety- and mission-critical applications by bridging the gap between high-level objectives and formal, executable specifications. The proposed method, Robotic System Task to Model Transformation Methodology (RSTM2) is an ontology-driven, hierarchical approach using stochastic timed Petri nets with resources, enabling Monte Carlo simulations at mission, system, and subsystem levels. A hypothetical case study demonstrates how the RSTM2 method supports architectural trades, resource allocation, and performance analysis under uncertainty. Ontological concepts further enable explainable AI-based assistants, facilitating fully autonomous specification synthesis. The methodology offers particular benefits to complex multi-robot systems, such as the NASA CADRE mission, representing decentralized, resource-aware, and adaptive autonomous systems of the future.

💡 Research Summary

The paper introduces a novel methodology called RSTM2 (Robotic System Task to Model Transformation Methodology) that aims to close the “specification gap” between high‑level mission objectives and executable, formally verified robotic system specifications. The authors argue that current model‑based engineering tools either remain too informal (e.g., SysML) or require extensive manual effort to integrate environmental models, resources, and constraints, especially for safety‑critical, multi‑robot missions. To address this, RSTM2 combines three pillars: (1) a robotics‑specific ontology derived from IEEE 1872.2‑2021 (system‑environment interaction) and IEEE 1872.1‑2024 (task‑plan‑resource) standards, (2) an extended stochastic timed Petri net (PN) formalism that incorporates time, probabilistic outcomes, and continuous resource consumption, and (3) an automated pipeline that translates ontology‑driven parameters into a domain‑specific language (RSSL2) and finally into ROS 2 C++ code.

The ontology defines a compact set of concepts—Entity, Objective, Environment, State, Impact, Task, Plan, Action, Resource, Capability, Affordance, Interaction—each of which can be instantiated at three abstraction levels: Mission, System, and Subsystem. By reusing the same vocabulary across levels, the methodology ensures semantic consistency while allowing granularity to increase as the design moves from mission‑wide intent down to concrete actuator commands.

In the “System Specification Synthesis” (3S) phase, designers input a high‑level objective; the ontology automatically derives the necessary entities, required capabilities, and associated resources. These parameters are then mapped onto an extended stochastic timed PN. The PN extends the classic place‑transition model with timed transitions whose firing durations are sampled from user‑specified probability distributions (e.g., normal, uniform). Tokens represent logical conditions, while resources such as energy or memory are modeled as non‑negative real values attached to places. Inhibitor arcs enable conditional disabling of transitions, and concurrent transition instances can be scheduled according to priority, random, or probabilistic policies. This representation captures both the logical flow of actions and the quantitative consumption of limited resources, making it possible to evaluate feasibility, parallelism, and reliability analytically or via simulation.

The second phase, “RSSL2 Specification Generation,” uses a VS Code plug‑in with parameterized templates to produce a textual specification in the RSSL2 language. A custom RSSL2 compiler then translates this specification into distributed ROS 2 nodes, automatically generating topics for inter‑node communication. Consequently, the entire design—from mission concept to executable code—can be produced with minimal manual coding.

Monte Carlo simulation is employed to assess the stochastic PN model. By repeatedly sampling transition durations and resource consumptions, the framework estimates mission success probabilities, identifies bottlenecks, and quantifies the impact of resource depletion or component failures. Designers can explore alternative architectures (e.g., adding redundant sensors, increasing the number of robots) and instantly see how these changes affect performance metrics. The same PN can also be instantiated at runtime to monitor actual resource levels and trigger contingency plans when capabilities degrade.

A hypothetical case study illustrates the workflow. A team of robots must locate objects, pick them up, and build a vertical stack. At the mission level the objective is “stack N objects at location L.” The ontology derives a system‑level plan comprising navigation, perception, grasping, and stacking actions, each annotated with required capabilities (e.g., visual detection, force control) and resources (battery, bandwidth). Subsystem models detail sensor sampling rates, actuator power draws, and communication latencies, all encoded in the stochastic PN. Monte Carlo runs reveal, for example, that with three robots and a 2 kWh battery pack the probability of completing the stack within the time budget is 78 %, whereas adding a fourth robot raises it to 92 % at the cost of higher energy consumption. These insights directly inform design decisions.

The authors also discuss an explainable AI assistant that can query the ontology in natural language, suggest modifications (e.g., “replace the gripper with a suction cup to increase grasp success”), and automatically regenerate the 3S parameters and ROS 2 code. This demonstrates how the ontology not only structures the design but also enables human‑AI co‑creation.

In conclusion, RSTM2 offers a comprehensive, ontology‑driven pipeline that transforms mission goals into formally verified, resource‑aware stochastic models and finally into executable ROS 2 controllers. By integrating formal verification, uncertainty quantification, and automated code generation, the methodology promises to improve safety, reliability, and development efficiency for complex, multi‑robot missions such as NASA’s CADRE project.