Successful operation of life support systems for human exploration of space will depend largely on in situ sensors and integration of these sensors to an autonomous control architecture. The complexity of a life support system demands that these sensors be uniquely qualified for a number of specific applications. Such applications may require multifunctional monitoring of the sensor environment and intelligent interaction with other sensors and control elements. In situ sensors found in the chemical processing industry are largely not multifunctional and not necessarily selected for minimal weight, volume and power demand. Therefore, the state-of-the-art for in situ sensors found in the chemical processing industries is not directly transferrable to life support applications. Highly autonomous life support systems require reliable chemical, physical and biosensors requiring little maintenance and which are highly fault-tolerant. These sensors will operate in a hierarchical network involving distributed databases, prediction models, rule-based reasoning modules for hypothesis generation and verification, controllers and final control elements. This paper discusses the limitations of the state-of-the-art for in situ sensors and argues in favor of a program of adaptation and enhancement of off-the-shelf sensor technologies and of research and innovation to develop more appropriate sensor technologies for life support systems. This paper also discusses briefly the desirable characteristics of smart sensors for life support applications and presents some preliminary concepts for hierarchical integration of in situ sensors and control elements.