Learning Actionable Information Landscapes for Multimodal Active Sensing in Hawkmoths
Authors: Abdelrahman Sharafeldin, Yaqing Wang, Simon Sponberg, Hannah Choi
Organizations: Machine Learning Center, Georgia Institute of Technology, Atlanta, GA 30332 · School of Physics, Georgia Institute of Technology, Atlanta, GA 30332 · School of Mathematics, Georgia Institute of Technology, Atlanta, GA 30332
Animals integrate information across sensory modalities to guide exploration, suggesting that they maintain internal estimates of where actions are expected to reduce uncertainty. However, how such action-conditioned information maps are learned from multimodal sensory experience remains unclear, as previous studies have been largely confined to unimodal sensing frameworks. We introduce a multimodal predictive-coding framework for learning uncertainty-reduction maps from visual and mechanosensory observations in hawkmoth flower interaction tasks. A variational generative model trained on simulated environments recovers modality-specific information landscapes over the action space, capturing how visual patterns, surface geometry, and mechanosensory cues shape the value of exploratory actions. These learned maps are consistent with several features of hawkmoth behavior, including nonzero tracking offsets, probing along visual patterns, and geometry-dependent exploration. Applying the model to behavioral data from flower tracking and nectary search, we find that natural trajectories tend to occupy regions of high uncertainty reduction in these learned maps. Building on these observations, we develop an uncertainty-gated lexicographic algorithm that switches between information gathering and reward exploitation based on perceptual uncertainty. In nectary search, uncertainty gating improves learning relative to exploit-only, ε-greedy, fixed-time, and uncertainty-calibrated non-lexicographic controls, with better generalization and robustness to sensory and curvature shifts. These results link information landscapes learned through generative modeling to biological sensing and embodied exploration.
Figure 3: Uncertainty-reduction maps in nectary search. (a) Mechanosensory and visual UR maps across trajectories and flower conditions. (b) Mean map UR along radial and circular trajectories. (c) Angular anisotropy of final UR maps: variance of mean UR across 16 equal angular sectors around the flower center, weighted by pixel counts and normalized by total map variance. (d) Modality weights along an example trajectory and mean mechanosensory weight with and without contact. (e) Mean visual, mechanosensory, weighted-unimodal, and aggregate-entropy UR across flower curvatures. (f) Realized information gain along biological and matched random trajectories. Shading and error bars in (b–e) show SEM. In (f), ∗∗p<0.01 (t-tests).
Figure 4: UR in flower tracking. (a) Action-conditioned UR map with flower and moth trajectories. (b,d) Mean realized UR for D. elpenor and M. stellatarum at medium and high brightness, corresponding to different levels of visual noise. (c) Biological tracking compared with exact flower tracking and shuffled moth trajectories. ∗∗p<0.01 (t-tests).
Figure 5: Aggregate-entropy control and baselines. (a) Training success rate, mean episode length, and posterior uncertainty across flower curvatures and visual patterns (icons). Curves show 25-epoch moving averages; shading denotes SEM across independent training seeds. (b) Validation success rates (mean ± SEM) under curvature changes and sensory degradation for cross-pattern flowers.
Appendix figures & tables6 assets
Supplementary material from the paper’s appendix.
Appendix
Figure S1: Example training trajectories from the realistic trajectory sampler
Figure S2: Example behavioral trajectories from the Deora et al. [ 4 ] dataset.
Figure S3: Additional training metrics corresponding to Figure 5 . (a) Mean aggregate-entropy reduction per exploration step for adaptive and fixed-time switching. Only these baselines are plotted here since they use the exploration head. (b) Episode return. (c) Exploration fraction. Shading denotes SEM across independent training runs.
Figure S4: Two successful aggregate-entropy control episodes on a funnel flower with a cross pattern. Left: head (circles) and proboscis-tip (triangles) trajectories, colored purple during exploration and green during exploitation. Right: smoothed aggregate entropy (orange), selected mode (blue), and switching thresholds (dashed), with matching phase shading. The examples switch at steps 11 and 8 and reach the nectary at steps 16 and 21, respectively; steps are indexed from zero.
Environment
Gate
Hysteresis
Fixed step
η
Funnel + cross
530
60
33
0.07
Funnel
535
60
40
0.05
Flat + cross
520
50
40
0.0065
Flat
550
52
10
0.004
Appendix
Table 1: Settings for the control experiments in Figure 5 .
Figure S5: Weighted-unimodal control. (a) Training success and episode length in four flower conditions. (b) Curvature and sensory-shift validation after funnel-cross training.
Active sensing links behavior and learning through an action-perception loop: actions determine the observations used to update internal predictive models of perception, which subsequently guide the next actions. Predictive-coding frameworks provide a natural way to model this process, since internal representations are continuously updated to predict future observations. Here, we ask how exploratory and exploitative behavioral strategies shape these internal predictive representations. We build an online learning agent in a tree-like maze with a controllable parameter regulating the balance between exploratory and exploitative regimes. The agent updates a predictive-coding-based perception model from experience generated by its own behavior. The model predicts both future maze states and reward probability, allowing the agent to select actions either by expected information gain during exploration or by predicted reward during exploitation. We show that the resulting internal predictive representations depend strongly on the agent's behavioral regime. Exploratory agents develop representations that are more spatially organized and better preserve the structure of maze transitions in latent space. In contrast, exploitative agents learn less organized representations. We then train this predictive model on natural trajectories of water-deprived mice navigating the same maze and compare the resulting representations with those learned from agent trajectories. More exploratory mice show representational geometries that closely match those of exploratory agents, whereas mice with more restricted visitation patterns resemble reward-driven, exploitative agents. Together, these findings suggest that exploration enables predictive models to form generalized internal representations by organizing latent space around both spatial location and transition context in artificial agents and animals.
Perception in biological systems is inherently noisy, requiring organisms to make decisions under uncertainty where misclassification can be costly or fatal. We present an Artificial Life predator-prey model of foraging under noisy perception, and compare agent performance when using various policies that take into account their noisy predictions. Through controlled experiments under both symmetric and asymmetric perceptual noise, we show that blindly trusting perceptual labels leads to catastrophic failure as noise increases, while uncertainty-aware strategies significantly improve survival and reduce fatal errors. We further observe qualitative regime shifts in behaviour, with agents transitioning from exploratory to conservative strategies as uncertainty increases. Our model links risk-sensitive foraging, ecological information use, and Artificial Life by showing that explicit information gathering can improve robustness when perception is unreliable. These results highlight the importance of uncertainty-aware decision-making and provide an interpretable artificial life analogue to robust learning with noisy labels.
Accurate action models are critical for effective planning. Existing action-model learning methods largely assume simple action representations or become computationally intractable when learning conditional and quantified effects. We present Online Hypothesis-Driven Conditional Action Model Learning (OHCAM), an online approach for learning action models with conditional and quantified effects from limited interactions with the environment. OHCAM maintains a belief over hypothesized action models and actively selects informative actions to reduce uncertainty by maximizing disagreement among competing hypotheses, while being robust to noisy observations. To enable scalability, OHCAM begins with a small set of simple action model hypotheses and expands to more complex conditions only when the current hypotheses become inconsistent with the data. Experiments on six benchmark planning domains demonstrate that OHCAM is sample efficient in learning action models that solve substantially more tasks than baselines, even with observation noise. We validate OHCAM on two tasks using a Kinova Gen3 robot, demonstrating the real-world applicability of our approach.
Jeffrey Jewett, William Solow, Sandhya Saisubramanian