Autonomous tracking of dynamic targets by unmanned ground vehicles (UGVs) is a crucial problem in various applications. However, existing approaches often suffer from low computational efficiency, inadequate handling of target visibility and limited adaptability to the rapid motion of the target. To address these limitations, this paper proposes a planner that combines target prediction-based topological path generation with a spatial-temporal trajectory optimization method. The future topological paths of target are predicted via motion primitives, and local topological search generates multiple candidate tracking paths for the tracker. Then an optimal path is selected on the basis of visibility cost, energy cost, and topological consistency. The spatial-temporal optimization method generates a smooth, dynamically feasible trajectory with minimum squared jerk and time. Considering both tracking distance and visibility constraints, the effectiveness of the proposed method we validate is through various simulation experiments.

Adapting billion-parameter language models to a downstream task is still costly, even with parameter-efficient fine-tuning (PEFT). We re-cast task adaptation as output-distribution alignment: the objective is to steer the output distribution toward the task distribution directly during decoding rather than indirectly through weight updates. Building on this view, we introduce Steering Vector Decoding (SVDecode), a lightweight, PEFT-compatible, and theoretically grounded method. We start with a short warm-start fine-tune and extract a task-aware steering vector from the Kullback-Leibler (KL) divergence gradient between the output distribution of the warm-started and pre-trained models. This steering vector is then used to guide the decoding process to steer the model's output distribution towards the task distribution. We theoretically prove that SVDecode is first-order equivalent to the gradient step of full fine-tuning and derive a globally optimal solution for the strength of the steering vector. Across three tasks and nine benchmarks, SVDecode paired with four standard PEFT methods improves multiple-choice accuracy by up to 5 percentage points and open-ended truthfulness by 2 percentage points, with similar gains (1-2 percentage points) on commonsense datasets without adding trainable parameters beyond the PEFT adapter. SVDecode thus offers a lightweight, theoretically grounded path to stronger task adaptation for large language models.