ESPADA: Execution Speedup via Semantics Aware Demonstration Data Downsampling for Imitation Learning

Behavior-cloning based visuomotor policies enable precise manipulation but often inherit the slow, cautious tempo of human demonstrations, limiting practical deployment. However, prior studies on acceleration methods mainly rely on statistical or heuristic cues that ignore task semantics and can fail across diverse manipulation settings. We present ESPADA, a semantic and spatially aware framework that segments demonstrations using a VLM-LLM pipeline with 3D gripper-object relations, enabling aggressive downsampling only in non-critical segments while preserving precision-critical phases, without requiring extra data or architectural modifications, or any form of retraining. To scale from a single annotated episode to the full dataset, ESPADA propagates segment labels via Dynamic Time Warping (DTW) on dynamics-only features. Across both simulation and real-world experiments with ACT and DP baselines, ESPADA achieves approximately a 2x speed-up while maintaining success rates, narrowing the gap between human demonstrations and efficient robot control.

A Study on Quantifying Sim2Real Image Gap in Autonomous Driving Simulations Using Lane Segmentation Attention Map Similarity

Autonomous driving simulations require highly realistic images. Our preliminary study found that when the CARLA Simulator image was made more like reality by using DCLGAN, the performance of the lane recognition model improved to levels comparable to real-world driving. It was also confirmed that the vehicle's ability to return to the center of the lane after deviating from it improved significantly. However, there is currently no agreed-upon metric for quantitatively evaluating the realism of simulation images. To address this issue, based on the idea that FID (Fr\'echet Inception Distance) measures the feature vector distribution distance using a pre-trained model, this paper proposes a metric that measures the similarity of simulation road images using the attention map from the self-attention distillation process of ENet-SAD. Finally, this paper verified the suitability of the measurement method by applying it to the image of the CARLA map that implemented a realworld autonomous driving test road.

Effects of Sim2Real Image Translation on Lane Keeping Assist System in CARLA Simulator

Autonomous vehicle simulation has the advantage of testing algorithms in various environment variables and scenarios without wasting time and resources, however, there is a visual gap with the real-world. In this paper, we trained DCLGAN to realistically convert the image of the CARLA simulator and evaluated the effect of the Sim2Real conversion focusing on the LKAS (Lane Keeping Assist System) algorithm. In order to avoid the case where the lane is translated distortedly by DCLGAN, we found the optimal training hyperparameter using FSIM (feature-similarity). After training, we built a system that connected the DCLGAN model with CARLA and AV in real-time. Then, we collected data (e.g. images, GPS) and analyzed them using the following four methods. First, image reality was measured with FID, which we verified quantitatively reflects the lane characteristics. CARLA images that passed through DCLGAN had smaller FID values than the original images. Second, lane segmentation accuracy through ENet-SAD was improved by DCLGAN. Third, in the curved route, the case of using DCLGAN drove closer to the center of the lane and had a high success rate. Lastly, in the straight route, DCLGAN improved lane restoring ability after deviating from the center of the lane as much as in reality.