Pattern Database (PDB) heuristics are an established approach in optimal classical planning that is used in state-of-the-art planning systems. PDBs are based on projections, which induce an abstraction of the original problem. Computing all cheapest plans in the abstraction yields an admissible heuristic. Despite their success, PDBs have only recently been adapted to numeric planning, which extends classical planning with numeric state variables. The difficulty in supporting numeric variables is that the induced abstractions, in contrast to classical planning, are generally infinite. Thus, they cannot be explored exhaustively to compute a heuristic. The foundational work that introduced numeric PDBs employed a simple approach that computes only a finite part of the abstraction. We analyze this framework and identify cases where it necessarily results in an uninformed heuristic. We propose several improvements over the basic variant of numeric PDBs that lead to enhanced heuristic accuracy.

While transformers excel in many settings, their application in the field of automated planning is limited. Prior work like PlanGPT, a state-of-the-art decoder-only transformer, struggles with extrapolation from easy to hard planning problems. This in turn stems from problem symmetries: planning tasks can be represented with arbitrary variable names that carry no meaning beyond being identifiers. This causes a combinatorial explosion of equivalent representations that pure transformers cannot efficiently learn from. We propose a novel contrastive learning objective to make transformers symmetry-aware and thereby compensate for their lack of inductive bias. Combining this with architectural improvements, we show that transformers can be efficiently trained for either plan-generation or heuristic-prediction. Our results across multiple planning domains demonstrate that our symmetry-aware training effectively and efficiently addresses the limitations of PlanGPT.

In this paper we exemplarily outline the application of a large scale tactile sensor to provide tactile perception to a mobile robot. The tactile sensors used within this project1are composed of piezo-resistive composite material sandwiched between two electrodes and arranged as sensor array. Based upon this approach, large 3D-shaped sensor covers that could be directly applied to the robots geometry were created. To simplify interpretation of sensor data, two major modes of operation have been implemented for the mobile robot: In autonomous mode, the tactile sensor acts as a safety collision sensor. In interaction mode touch and gestures on the robot surface are interpreted as interaction metaphors which allow intuitive human robot interaction. The sensor setup and integration will be described and the results of the experimental evaluation will be outlined.

In this contribution we present an innovative and trendsetting solution for safeguarding human-robot cooperative workplaces with high-payload robots through a combination of safeguarding technologies addressing both hard1- and soft2- safety considerations. This consists of a tactile floor with spatial resolution as a hard- safety sensor for workspace monitoring together with a projection system as a soft- safety component to visualize the boundaries of the safety zones. This safety concept is capable of establishing both manually defined safety zones and dynamically generated safety zones that are based on the current robot's joint positions and velocities, thus offering a maximum of free space around the robot to the user. The paper aims on introducing the novel safety concept and will briefly describe the development of the underlying technologies.

Human-robot interaction in a shared workspace permits and often even requires physical contact between humans and robots. A key technology to ensure that physical human robot interaction is safe is to monitor contact forces by providing the robot with a tactile sensor as an artificial skin.