Aerial search and rescue (SAR) rotorcrafts currently need multiple specialist human operators, increasing cost and the risk of downtime due to crew unavailability and mental stress. Autonomy can aid fewer operators performing multiple tasks, but the human operator must maintain situation awareness (SA) of crucial autonomous decisions. A key challenge is the cognitive stress on a multi-tasking human-in-the-loop (HITL) due to the AI agent making decisions without human understanding. Explainable AI (XAI) has often been proposed as a way to explain autonomy decisions, but current XAI solutions doesn’t adapt to real-time human factors in high stress and high stakes situations. Here, we allow an AI agent to perform autonomous rotorcraft navigation, whilst the HITL operator has to perform two simultaneous tasks: (i) search for a target on the ground by toggling an onboard camera, and (ii) maintain SA of the autonomous navigation task through our novel XAI interface. Our novel XAI approach leverages on dimensionality reduction techniques to visualize the reinforcement learning (RL) navigation’s internal states, highlighting patterns in its decision-making process through intuitive interactive clustering on saliency maps. To ensure convergence on performance, we design a two-way interface that allows the human to interpret AI decisions and then give feedback via a Large Language Model to modify the autonomous navigation. Testing demonstrates increased task performance (+43%), while experiencing substantial human reductions in physical demand (-53%), time pressure (-30%), effort (-23%), and frustration (-26%), but at the cost of slightly increased mental demand (+12%).

The intrinsic predictability of a given time series indicates how well an (ideal) algorithm could potentially predict it when trained on the time series data. Being able to compute the intrinsic predictability helps the developers of prediction algorithms immensely in deciding whether there is further optimization potential, as it tells them how close they are to what is (theoretically) achievable. We call the intrinsic predictability the predictability upper bound ¶imax and propose a novel method for quantifying and estimating it for univariate numeric time series. So far, this has only been done for symbolic time series, even though most real-world time series are numeric by nature. We base our technique on the close relationship between entropy and predictability, utilizing the entropy rate of a time series to compute ¶imax . Since existing entropy rate estimators, such as those based on the Lempel-Ziv compression algorithm, only work for symbolic data, we develop new estimators using tolerance thresholds for matching numeric values. We demonstrate that ¶imax is an effective upper bound that characterizes the intrinsic predictability of a time series. We give formal proofs and we validate our arguments experimentally by comparing ¶imax with the prediction accuracy of different state-of-the-art models on various real-world datasets from different domains.