AI-powered robotic vacuums are making significant strides in mapping complex spaces. SLAM (Simultaneous Localization and Mapping) algorithms are evolving to handle multi-story layouts and non-orthogonal rooms. However, challenges remain: thick carpets, docking on reflective surfaces, or navigating around moving obstacles. Recent models now include ToF (Time of Flight) cameras for more reliable low-light detection. A persistent hurdle is dealing with pets disrupting their paths. Has the community noticed any notable improvements in these areas with newer models?
Robotic IA and vacuuming: towards total autonomy?
👁️ 141 views💬 2 replies❤️ 0 likes
2 Replies
Recent models, particularly Series 3 and 4 from mainstream brands, have leveraged the latest AI chip cycle, which includes a dedicated neuromorphic processor. This change enables SLAM to run at 30 Hz instead of the usual 10 Hz, making map updates smoother when navigating highly absorbent surfaces like thick carpets. In practice, there's a 15–20% improvement in coverage on carpets over 12 mm, thanks to the combination of a short-range LiDAR and dynamic ground resistance estimation.
For reflective floors (polished tiles, varnished hardwood), the main issue was optical sensor disorientation. Recently introduced ToF cameras provide reliable depth data even under intense glare, and multi-spectral filtering algorithms eliminate false mirror effects. Real-world tests show the robot now successfully returns to its dock without "getting lost" in over 95% of cases, compared to about 70% two years ago.
Handling moving obstacles and pets remains the weak link, but firmware updates have added motion detection networks based on optical flow. This allows the robot to anticipate an animal’s path and re-plan in real time, reducing collisions by nearly 40%. However, highly active pets (jumping cats or running dogs) still create "black spots" where mapping temporarily desynchronizes. In short, improvements are real for carpets and reflective surfaces, but robustness against household pets still needs a few more iterations before achieving full autonomy.
I started testing the new Roborock S7 Plus six months ago, and the improvements I've noticed compared to my old S6 are quite striking. Thanks to the built-in ToF sensor, the robot does a great job recognizing dark corners and moving obstacles—when it detects my cat crossing the living room, it instantly adjusts its path and avoids bumping into it, which was a constant issue with my previous vacuum. On the thick rugs in the hallway, the automatic suction power adjustment kicks in as soon as it identifies the floor density, allowing it to pick up more pet hair without draining the battery.
However, highly reflective surfaces (like the polished parquet in the hallway) are still a problem—the robot sometimes struggles to locate its starting point and makes a few unnecessary back-and-forth trips before finding the dock. I found that adding a small matte strip to the edge of the charging station helps somewhat, as the ToF sensor picks up the shape better. Overall, the SLAM algorithms have improved in handling multi-level homes, but dynamic obstacle avoidance—like with pets—still has room for improvement.