Path planning is a fundamental issue in the aspect of robot navigation. As robots work in 3D environments, it is meaningful to study 3D path planning. To solve general problems of easily falling into local optimum and long search times in 3D path planning based on the ant colony algorithm, we proposed an improved the pheromone update and a heuristic function by introducing a safety value. We also designed two methods to calculate safety values. Concerning the path search, we designed a search mode combining the plane and visual fields and limited the search range of the robot. With regard to the deadlock problem, we adopted a 3D deadlock-free mechanism to enable ants to get out of the predicaments. With respect to simulations, we used a number of 3D terrains to carry out simulations and set different starting and end points in each terrain under the same external settings. According to the results of the improved ant colony algorithm and the basic ant colony algorithm, paths planned by the improved ant colony algorithm can effectively avoid obstacles, and their trajectories are smoother than that of the basic ant colony algorithm. The shortest path length is reduced by 8.164%, on average, compared with the results of the basic ant colony algorithm. We also compared the results of two methods for calculating safety values under the same terrain and external settings. Results show that by calculating the safety value in the environmental modeling stage in advance, and invoking the safety value directly in the path planning stage, the average running time is reduced by 91.56%, compared with calculating the safety value while path planning.

Differentiable architecture search (DARTS) transforms architectural optimization into a super-network optimization by stacking two cells (2 c.). However, repeatedly stacking two cells is a sub-optimal operation since cells in different depths should be various. Besides, we find that the performance is slightly improved by increasing the number of searched cells (e.g., from 2 c. to 5 c.), but it leads to uneven resource allocation. This letter proposes a regularized DARTS (RDARTS) to adjust the architectural differences and balance degrees of freedom and resource allocation. Specifically, we use separate architectural parameters for two reduction cells and three normal cells, and then propose a Reg distance to calculate the difference between cells. We design a new validation loss which is the weighting of cross-entropy and Reg loss, and introduce an adaptive adjustment method. Results show that RDARTS achieves the top-1 accuracy of 97.64% and 75.8% on CIFAR and ImageNet.

Differentiable neural architecture search (NAS) methods have achieved comparable results for low search costs and high performance. Existing differentiable methods focus on searching microstructures in micro space, lacking in exploring macrostructures. However, different networks should have different macrostructures rather than all evenly distributed. This article proposes an M ² NAS optimization system to optimize macrostructure and microstructure jointly. Specifically, we initialize a network with maximum complexity, where down-samplings occur at the beginning. Then, iteratively optimize the macrostructure and microstructure. For macro search, we establish a macro space and explore this space by generating candidates and initializing weights and architectural parameters for candidate networks, i.e., network transmission. For micro search, we propose a progressive pruning method to eliminate the vast quantization error caused by one-time pruning. With the number of parameters decreasing during the search, M ² NAS obtains a series of networks with different complexity through network selection, forming a Pareto-optimal set. Results show that networks obtained by M ² NAS have a better tradeoff between accuracy and complexity. A network searched on ImageNet achieves 77.4% and 76.3% top-1 accuracy, with and without data augment during retraining. When taking different pretrained models as backbones and combining them with Faster RCNN on COCO, our model can get 35.6% $(AP)$ and 55.7% $(AP_{50})$ , higher than typical NAS models, second only to ResNet-50 results.

Neural Architecture Search (NAS) is an emerging topic in machine learning and computer vision. The fundamental ideology of NAS is using an automatic mechanism to replace manual designs for exploring powerful network architectures. One of the key factors of NAS is to scale-up the search space, e.g., increasing the number of operators, so that more possibilities are covered, but existing search algorithms often get lost in a large number of operators. For avoiding huge computing and competition among similar operators in the same pool, this paper presents a scalable algorithm by factorizing a large set of candidate operators into smaller subspaces. As a practical example, this allows us to search for effective activation functions along with the regular operators including convolution, pooling, skip-connect, etc. With a small increase in search costs and no extra costs in re-training, we find interesting architectures that were not explored before, and achieve state-of-the-art performance on CIFAR10 and ImageNet, two standard image classification benchmarks.