We propose a framework to enable multipurpose assistive mobile robots to autonomously wipe tables to clean spills and crumbs. This problem is challenging, as it requires planning wiping actions while reasoning over uncertain latent dynamics of crumbs and spills captured via high-dimensional visual observations. Simultaneously, we must guarantee constraints satisfaction to enable safe deployment in unstructured cluttered environments. To tackle this problem, we first propose a stochastic differential equation to model crumbs and spill dynamics and absorption with a robot wiper. Using this model, we train a vision-based policy for planning wiping actions in simulation using reinforcement learning (RL). To enable zero-shot sim-to-real deployment, we dovetail the RL policy with a whole-body trajectory optimization framework to compute base and arm joint trajectories that execute the desired wiping motions while guaranteeing constraints satisfaction. We extensively validate our approach in simulation and on hardware. Video: https://youtu.be/inORKP4F3EI

The opacity of neural networks leads their vulnerability to backdoor attacks, where hidden attention of infected neurons is triggered to override normal predictions to the attacker-chosen ones. In this paper, we propose a novel backdoor defense method to mark and purify the infected neurons in the backdoored neural networks. Specifically, we first define a new metric, called benign salience. By combining the first-order gradient to retain the connections between neurons, benign salience can identify the infected neurons with higher accuracy than the commonly used metric in backdoor defense. Then, a new Adaptive Regularization (AR) mechanism is proposed to assist in purifying these identified infected neurons via fine-tuning. Due to the ability to adapt to different magnitudes of parameters, AR can provide faster and more stable convergence than the common regularization mechanism in neuron purifying. Extensive experimental results demonstrate that our method can erase the backdoor in neural networks with negligible performance degradation. Benefited from the powerful representation learning ability, neural networks (NNs) play an imperative role in many fields, especially for image processing Al-Saffar et al. (2017). However, the brilliant feat of NN also makes it a focal point of many attacks, one of the most threatening among which is the backdoor attack Gu et al. (2017); Liu et al. (2017); Barni et al. (2019); Chen et al. (2017). By mixing poisoned data into the training set, backdoor attack can control the victim NN to output attacker-chosen predictions for triggered inputs, while hardly disturb the predictions of normal inputs.

Tesla has released a new software update with major improvements to its automatic wiper trained with a new deep neural net previously referred to as "Deep Rain." Like most premium vehicles today, Tesla has an automatic wiper system that automatically matches the speed of the wipers to the intensity of the rain or snow. Instead, the automaker is using its Autopilot cameras to feed its computer vision neural net to determine the speed for the wipers. It has been deployed in Tesla vehicles since last year, but some owners have been complaining that it is not as accurate as other systems using rain sensors. Lately, CEO Elon Musk has been talking about Tesla releasing a new "Deep Rain" neural net to improve the automatic wipers.

Technology behind self-driving cars has come a long way in the last several years, racing from the lab to real-life city streets, but for the artificially-intelligent vehicles, obstacles, in this case the all-too-common pothole, still abound. Tesla CEO and avid Twitter user, Elon Musk, said that his company is working to change that. In a reply to multiple tweets regarding test runs of Tesla's new and more automated autopilot feature, Musk seems to suggest that his company is working on a solution to help the company's vehicles steer clear of dangerous potholes. For artificially-intelligent vehicles, obstacles, in this case the all-too-common pothole, still abound. Tesla introduced Autopilot in 2014.

These very different inventions share one thing in common: they were created by women. Despite their significant contributions, many of these female inventors have gone unrecognized. In honor of International Women's Day, take a moment to appreciate these six inventions we wouldn't have without women. Thanks to Melitta Bentz from Germany, you don't have to worry about grounds in your cup of joe. In 1908, Bentz was in search of a better coffee-drinking experience.

To say the $35,000 Model 3 is important to Tesla would be an understatement. Judging by its pre-orders (the highest the industry has seen), it's already the most anticipated car ever. It's the culmination of CEO Elon Musk's nearly 12-year-old "master plan" to bring an affordable EV to market. Now that the Model 3 is here I can positively say it's a joy to drive, but it's also frustrating to do some of the simplest things in the cockpit. During a single trip I went from an electric-motor fueled grin on my face to throwing my hands up in exasperation trying to adjust the cruise control follow distance.