Weed control is a major task in farming, it is critical to ensure yield and profit. Challenges are the sustainability and the lack of work force. Automation, particular agriculture robots may contribute to ensure yield and gain sustainability. This work investigates mobile manipulation for mechanical weeding in agriculture, addressing relevant techniques leading towards a robotic solution for weed control. The main topics are vision-based positioning and the design of mechanical tools for weed control in agriculture. For positioning of the tool, an eye-in-hand visual servoing approach is used. Computer vision is enhanced by a FPGA implementation, resulting in a fast working rate and a high precision within the unstructured agriculture field environment. Based on the millimeter accuracy of the positioning, new fast acting weeding tools were developed, which are able to treat small single plants. The performance is demonstrated in field and lab experiments.weiterlesen