Pervasive computing is rapidly achieving wide spread acceptance as a computing paradigm. This paradigm has moved from a once visionary concept to what is now a tangible solution which is being deployed in many domains. Recent technological improvements including increased computing power, low cost sensing and actuating technology, and increased usage of mobile devices along with the widespread availability of both wired and wireless networks have all contributed to this success. Nevertheless, there are still a number of challenges and open issues which remain unresolved. A central feature of intelligent environments which are facilitated through pervasive computing is the capability to perform automated activity recognition. Before an intelligent environment can provide useful and appropriate levels of automated support an appreciation of the inhabitant’s ongoing activities is required. Activity recognition is an active area of research which has progressed from image and video based processing systems to those which make use of an array of heterogeneous sensors, seamlessly embedded within the environment. Although contemporarymachine learning techniques have been successfully appliedwithin this domain, the ability of an activity recognition system tomanage the diversity and uniqueness of all possible scenarios lies at the heart of the challenge in producing a truly scalable solution. In an effort to address this challenge a trend has been adopted which has moved towards using domain knowledge such as user activity profiles, domain-specific constraints and common sense heuristics as the basis for the learning and development of a new generation of activity recognition models. This Special Issue on Knowledge Driven Activity Recognition in Intelligent Environments presents a collection of relevant work within this area which has made significant advances to overcome a number of the aforementioned challenges. In total, seven papers are included which cover topics ranging from recognition of activities based on information gathered from wearable sensors to ontological based approaches used to capture domain knowledge. In the work presented by Wang et al. entitled ‘‘Recognizing multiple-user activities using wearable sensors in a smart home’’, the authors address the classical problem of activity recognition in cases of multiple residents. At the core of the work is a wearable platform which collected information using an array of multi-modal sensors that relates the users’ location, their interaction with objects, and human-to-human interactions along with a collection of environmental factors. Two temporal probabilistic models, namely a Coupled Hidden Markov Model and a Factorial Conditional Random Field, are evaluated for their ability to recognize activities from this data. Data was collected and evaluated for two users’ behaviours within an intelligent environment over a 10 day period. The results demonstrate the effectiveness of the approaches for detecting multiple-user activities. Nevertheless, the work also demonstrates that within the myriad of available sensor recordings, consideration of a smaller subset of features had the potential to further improve the recognition process. The use of Partially observable Markov decision process (POMDP) models is presented in the work by Hoey et al. in their paper, ‘‘Rapid specification and automated generation of prompting systems to assist people with dementia’’. In this work the authors introduce a knowledge driven method for automatic activity recognition and prompting using POMDPs. The concepts in thiswork have been demonstrated through evaluation of the approach in supporting the activity ofmaking a cup of tea within an intelligent kitchen environment. Detailed analysis of the results demonstrates the successful application of the approach. These experiments also highlight the ability of the approach to manage errors whichmay have existed within the sensor readings and with unpredicted user interactions. Tastan and Sukthankar, in their paper ‘‘Leveraging human behaviour models to predict paths in indoor environments’’, consider the important topic of path prediction and tracking within indoor environments. Experimental studies within the paper were conducted on a model based on visually-guided steering and were considered within the realms of both simulated and real data sets in an effort to convey the effectiveness of the proposed techniques. The analysis of the results demonstrated the superiority of the proposed techniques over standard planning models. Rashidi and Cook’swork addresses the notion of transferring the knowledge gained in learned activities from one domain to support the development of the recognition processes required in new environments in their paper entitled ‘‘Activity knowledge transfer in smart environments’’. Such an approach offers many advantages in the development of activity recognition algorithms, specifically in terms of reducing the amount of data collection required along with benefitting from